This item is only available as the following downloads:
1 DEVELOPMENT OF VIRTUAL WORLDS IN AGRICULTURE AND NATURAL RESOURCES FOR SIMULATION EXPERIMENTS AND ELEARNING USING ONTOLOGY BASED SIMULATION By YUNCHUL JUNG A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Y unchul J ung
3 To my advisor, committee members, my parents and my fianc e
4 ACKNOWLEDGMENTS It has been a long journey since beginning my stud ies at the graduate school level. I always liked to pursuit new topics and to learn new technologies for the future, and I have finally reached the end of my journey I would like to since rely thank my advisor, Dr. Howard W. Beck. It was lucky for me to meet and collaborate with him. His ideas and advice have stimulated me to achieve my research objectives. His immense passion for research and intellectual insights were great motivating fac tors in my work. Secondly, I would also like to extend my appreciation to committee members for their encouragement and guidance to complete this work. Dr. Michael E. Bannister introduced the concept of the learning forest system, and I am very thankful t o him for directing the Virtual Learning Forest project sponsored by National Institute of Food and Agriculture (NIFA). Dr. Greg Kiker has guided me to understand systems based on natural resources and systematic approaches. Dr. Benjamin C. Lok has given v aluable advice by seeing the usability of my work with his enormous experiences in virtual world s Dr. Kelly T. Morgan has been a great help for guiding me in developing various practical simulations with valuable field experimental experience I would l ike to thank my friends, Chang Hwan, Chang Burm and Steve Shin at UF. Their family also gave me great pleasure of life in Gainesville and help me to feel the warmth of family. Many thanks go to my parents, my brother and his family. Finally, I take this op portunity to thank my fianc Kyungmi for her long lasting love, supports and sacrifices
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 12 ABSTRACT ................................ ................................ ................................ ................... 13 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 15 Statement of Problem ................................ ................................ ............................. 15 Limitations of VWE as an Educational Tool ................................ ...................... 16 Lack of Methods of VWE for Agriculture and Natural Resources ..................... 19 Appropriate Approach for Building Virtual Objects for VWEs ........................... 21 Specifi c Objectives ................................ ................................ ................................ 25 Approach ................................ ................................ ................................ ................ 25 Dissertation Layout ................................ ................................ ................................ 28 2 VIRTUAL LEA RNING FOREST: AN APPROACH FOR DEVELOPING EDUCATIONAL EXPERIMENTS AND SIMULATIONS WITHIN A VIRTUAL WORLD ENVIRONMENT ................................ ................................ ....................... 30 Background ................................ ................................ ................................ ............. 30 Virtual World Environment (VWE) ................................ ................................ ........... 31 Virtual Learning Forest (VLF) World ................................ ................................ ....... 32 Main Features of VLF ................................ ................................ ............................. 35 Scene Components for Virtual World ................................ ............................... 35 Tree object ................................ ................................ ................................ 35 Terrain object ................................ ................................ ............................. 37 Forest object ................................ ................................ .............................. 38 Instruments for Mensuration ................................ ................................ ............. 39 Samplin g area tool ................................ ................................ ..................... 39 Logger tape ................................ ................................ ................................ 40 Clinometer ................................ ................................ ................................ .. 41 Diameter tape ................................ ................................ ............................ 41 User Interface Objects ................................ ................................ ...................... 42 Welcome board ................................ ................................ .......................... 42 Tool bag ................................ ................................ ................................ ..... 43 Instruments guide ................................ ................................ ...................... 43 Logger note ................................ ................................ ................................ 43
6 Evaluation of VLF System ................................ ................................ ....................... 44 Evaluation of System during Development ................................ ....................... 44 Evaluation of System Performance ................................ ................................ .. 45 Evaluation of System Usa bility : System Design ................................ .............. 46 Evaluation of System Usability: Learning Efficiency ................................ ......... 48 Summary ................................ ................................ ................................ ................ 53 3 VIRTUAL WORLD DESIGN APPROACH FOR ABSTRACTING AND CHARACTERIZING BASIC VIRTUAL OBJECTS ................................ ................... 73 Background ................................ ................................ ................................ ............. 73 Complexities of Proprietary Model for VWE ................................ ............................ 74 Model Structure ................................ ................................ ................................ 74 Data Flow ................................ ................................ ................................ ......... 75 Graphic Model Format ................................ ................................ ...................... 77 Conceptualization of the VLF Educational Objects ................................ ................ 78 Domain Interests ................................ ................................ .............................. 79 Conceptualization ................................ ................................ ............................. 80 Model Representation ................................ ................................ ...................... 80 Abstraction of Objects into For m of Virtual Objects ................................ ................ 82 Tree Virtual Object ................................ ................................ ........................... 82 Clinometer Virtual Object ................................ ................................ .................. 84 Logger Tape Virtual Object ................................ ................................ ............... 85 Diameter Tape Virtual Object ................................ ................................ ........... 87 Sampling Area Virtual Objects : Center pole, Boundary tape, and Tree tag ..... 88 Summary ................................ ................................ ................................ ................ 90 4 AN ONTOLOGY BASED APPROACH TO IMPROVING MODEL INTEROPERABILITY ................................ ................................ ............................ 107 Background ................................ ................................ ................................ ........... 107 Model Interoperability ................................ ................................ ............................ 108 Ontology based Simulation Model Approach ................................ ................. 109 Exchangeable Digital Content File Format Approach ................................ ..... 110 Ontology based Model and System ................................ ................................ ...... 112 Ontology Design ................................ ................................ ............................. 113 Tree ontology ................................ ................................ ........................... 113 Sampling area tool ................................ ................................ ................... 115 Diameter tape ................................ ................................ .......................... 116 Logger tape ................................ ................................ .............................. 118 Clinometer ................................ ................................ ................................ 119 Ontology Implementation ................................ ................................ ................ 121 Authoring tools ................................ ................................ ......................... 122 Building an ontology ................................ ................................ ................. 123 Digital Content Generator ................................ ................................ ............... 127 Graphic model geometry generator ................................ ......................... 127 Graphic model module generator ................................ ............................. 130
7 Model Interoperability ................................ ................................ ..................... 131 Summary ................................ ................................ ................................ .............. 134 5 CONCLUSIONS, CONTRIBUTIONS, AND FUT URE DIRECTIONS .................... 157 Conclusions ................................ ................................ ................................ .......... 157 Presenting a Virtual World in the Agriculture and Natural Resource Domain 157 Abstracting and Characterizing Virtual Objects ................................ .............. 157 Methodology for Enhancing Model Interoperability by Using An Ontology ..... 158 Contributions ................................ ................................ ................................ ......... 159 Future Directions ................................ ................................ ................................ .. 159 Expansion of Model Interoperability ................................ ............................... 159 Evaluation for Contribution of VLF to Real Education ................................ .... 160 Digital Forest Learning Library ................................ ................................ ....... 160 Ontology Reasoning ................................ ................................ ....................... 160 APPENDIX: EVALUATION FORM OF VLF SYSTEM ................................ ................. 161 LIST OF REFERENCES ................................ ................................ ............................. 166 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 173
8 LIST OF TABLES Table page 2 1 List of average launching speeds of VLF and computer specifications (tested at July ~ Aug, 2011) ................................ ................................ ............................ 55 2 2 List of questions and scores of evaluation form: Question Category A ............... 56 2 3 List of questions and scores of evaluation form: Question Category B ............... 57 2 4 List of questions and scores of evaluation form: Question Category C .............. 58 2 5 List of actual and measured tree height and DBH at experiment sector A. ........ 58 2 6 List of actual and measured tree height and DBH at experiment sector B. ........ 59 2 7 Statistical values representing errors and measurement agreements .............. 59 3 1 Abstraction of Tree virtual object. ................................ ................................ ....... 91 3 2 Abstraction of Clinometer virtual object. ................................ ............................. 91 3 3 Abstraction of Logger Tape virtual object. ................................ .......................... 91 3 4 Abstraction of Diameter Tape virtual object. ................................ ....................... 92 3 5 Abstraction of Sampling Area Tool virtual object. ................................ ............... 92
9 LIST OF FIGURES Figure page 2 1 An example of conceptualization of forest learning system ............................... 60 2 2 An example of conceptualization of a forest learning system ............................ 61 2 3 Virtual Learning Forest (VLF) world Client/Server architecture showing independent and extendable relationships between VWE and developed virtual world elements ................................ ................................ ......................... 62 2 4 A virtual object of the sampling area tool ................................ ........................... 63 2 5 A virtual object of the logger tape ................................ ................................ ....... 6 4 2 6 A virtual object of the clinometer which is used to measure the height of tree .. 65 2 7 A virtual object of the diameter tape ................................ ................................ ... 66 2 8 The Welcome Board that is displayed at the start of each session. .................... 67 2 9 The Tool Bag of the instruments that are available for use in the VLF. .............. 68 2 10 A n user interface object of User Guide ................................ .............................. 69 2 11 The Logger Note used by a student to record tree measurements made with the instrumen ts. ................................ ................................ ................................ .. 70 2 12 An example of enhancement of user interface design and interaction: Diameter tool ................................ ................................ ................................ ...... 71 2 13 Locations of two experimen t sectors in the virtual forest ................................ ... 72 3 1 Structure of Classes which represent a forest virtual object following a proprietary method used for OpenWonderland format. ................................ ...... 93 3 2 Structure of Classes which represent a HUD object for instrument (diameter) following the required proprietary method used for OpenWonderland format. ... 93 3 3 A client architecture design for data centered representation of instrument objects ................................ ................................ ................................ ................ 94 3 4 A client architecture design for model centered representation of instrument objects ................................ ................................ ................................ ................ 94 3 5 A use case diagram of the Virtual Learning Forest. ................................ ............ 95
10 3 6 A sequence diagram of the virtual learning forest system, which focuses on the interaction and behaviors between actors and virtual objects. ..................... 96 3 7 A collaboration diagram of the virtual learning forest system. ............................ 97 3 8 Longleaf pine trees ................................ ................................ ............................. 98 3 9 A longleaf pine tree geometry model. ................................ ................................ 99 3 10 Level of detail of the longleaf pine consisting of 4 tree digital resolutions. ........ 100 3 11 Major parts of the clinometer (Suunto PM 5 model). ................................ ........ 100 3 12 Clinometer (Suunt o PM5 model) digital content ................................ ............... 101 3 13 Geometries of 3D clinometer digital content. ................................ .................... 102 3 14 Major parts of the logger tape and d etail of the ruler scale .............................. 102 3 15 Logger tape (Stanley Powerlock) geometry model ................................ ........... 103 3 16 Major parts of diameter tape (JIM GEM Pocket Diameter Tape) and drawing showing how to read a measurement ................................ .............................. 104 3 17 Diameter tape digital content ................................ ................................ ............ 104 3 18 Materials and tools used for assigning a sampling area ................................ .. 105 3 19 Sampling area tools (center pole, boundary tape, tree identification tag) ........ 106 4 1 A diagram of the longleaf pine model geometry ontology ................................ 136 4 2 A diagram of the longleaf pine model behavior ontology ................................ .. 137 4 3 A diagram of sampling area tool model geometry ontology .............................. 138 4 4 A diagram of sampling area tool model behavior ontology. .............................. 139 4 5 A diagram of diameter tape model geometry ontology ................................ ..... 140 4 6 A diagram of diameter tape model behavior ontology ................................ ...... 141 4 7 A diagram of logger tape model geometry ontology ................................ ......... 142 4 8 A diagram of logger tape model behavior ontology. ................................ ......... 143 4 9 Taxonomy of th e virtual object (clinometer) domain ontology. .......................... 144 4 10 A diagram of clinometer model geometry ontology ................................ ........... 145
11 4 11 A diagram of cl inometer model behavior ontology. ................................ ........... 146 4 12 A n implementation example of an instrument (clinometer) ontology within LyraBrowser. ................................ ................................ ................................ .... 147 4 13 VLF Digital Content class and its subclasses ................................ .................. 148 4 14 Taxonomical hierarchy of VLF DC Clinometer 3D Model class ....................... 148 4 15 Subclasses and associations of VLF Clinometer Shape class ........................ 149 4 16 Example of building a group of model behaviors for VLF instruments with the SimulationEditor, an ontology based simulati on authoring tool ....................... 150 4 17 Representation of Equation element class and their subclasses in the LyraBrowser ................................ ................................ ................................ .... 150 4 18 An equati on object for representing the tree height calculation with view angle in percentage and distance, in equation tab of EquationEditor .............. 151 4 19 An example of defining a symbol (SYMBOL Tree Heig ht) within Symbols tab in the EquationEditor ................................ ................................ ....................... 151 4 20 A n example of geometry instances in ontology representing a scale wheel part of an instrument (clinometer) within LyraBrowser. ................................ ..... 152 4 21 An example of generated COLLADA code for a scale wheel part of clinometer model ................................ ................................ ............................. 153 4 22 An example diagram of Clinometer model module generated from the ontolog ies ................................ ................................ ................................ ........ 154 4 23 A simple example of 3D model interoperability : Longleaf pine forest model. .... 155 4 24 A simple example of 3D model interoperability : Clinometer model. .................. 156
12 LIST OF ABBREVIATIONS ACM Association for Computing Machinery COLLADA Collaborative Design Activity DCC Digital Content Creation DEM Digital Elevati on Model FAO Food and Agriculture Organization GIS Geographical Information System HTML Hypertext Markup Language HUD Head Up Display IEEE Institute of Electrical and Electronics Engineers LOD Level of Detail SSM Soft System Methodology USDA U.S. Departmen t of Agriculture USGS U.S. Geological Survey VLF Virtual Learning Forest VR Virtual Reality VRML Virtual Reality Modeling Language VW Virtual World VWE Virtual World Environment WWW World Wide Web XML Extensible Markup Language
13 Abstract of Dissertatio n Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DEVELOPMENT OF VIRTUAL WORLDS IN AGRICULTURE AND NATURAL RESOURCES FOR SIMULATION EXPERIMENTS AND ELEARNING USING ONTOLOGY BASED SIMULATION By Yunchul Jung Dec ember 2013 Chair: Howard W. Beck Major: Agricultural and Biological Engineering The virtual world environment (VWE) is an emerging technology that offers advantages for experiencing vario us problems and exploring issues associated with diverse domains in a virtual world without disadvantages of the real world environments such as lack of access, expense, and in many cases potential hazards Although VWEs offer capabilities and potentials f or building a n agriculture virtual world, there are many problems related to creating agriculture digital content and sharing it in the domain. This research on development of virtual worlds in the agriculture and natural resource domain facilitate s develo pment of simulation experiments and education al tools by applying ontology based simulation methodology and resolv es some emerging problems on development of virtual world environments. To achieve this goal, a case study was developed in which basic virt ual objects in the agriculture and natural resource domain are characterized, and development methods are established for implementing digital content. A digital content ontology is constructed for storing the elements of digital content irrespective of gr aphic model formats. And, an ontology based simulation methodology is developed for the VWE to
14 realize behaviors of objects in real world. As a experimental example, a virtual world for e learning in forestry called the Virtual Learning Forest (VLF) is cre ated, and various instruments used for field laboratory based learning activities are developed. Generation engines for graphic model and digital content are constructed and utilized to create digital products automatically by cooperating with a model geom etry ontology and a model behavior ontology. An evaluation of system usability was conducted for 30 students in the VLF, and students gave slightly higher score than middle score (4 points in 7 scales) for the design of instruments. For the learning effi ciency, test scores shows that students developed their skill by increasing accuracy of measurement. T he result of the e valuation showed that the VLF system can enhance student experience of system usability, accessibility to digital content and expandabil ity of virtual world s in the domain. A model interoperability test was conducted by porting the content of the VLF into t w o different platforms: OpenSimulator ( http://opensimulator.org ) and OpenWonderland ( http://openwonderland.org ) I t is verified that the applied ontology based VWE methodology improves capability of sharing graphic models and object behavior and shows the possibility of building enhanced and expanded digi tal content. It will serve to facilitate sharing agriculture digital assets by improving model portability and interoperability between VWEs and contribute to decreased cost and complexity of building virtual worlds for simulation experiments and educatio n.
15 CHAPTER 1 INTRODUCTION The virtual world environment (VWE) is an emerging technology that offers advantages for experiencing various problems and exploring issues associated with diverse domains in a virtual world without disadvantages of the real wor ld environments such as lack of access, expense, and potential hazards Although VWEs offer capabilities and potentials for building a n agriculture virtual world, there are many problems related to creating agriculture digital content and sharing it in the domain. This research on development of virtual worlds in the agriculture and natural resource domain facilitate s development of simulation experiments and education al tools by applying ontology based simulation methodology and resolves some emerging pro blems on development of virtual world environments. The research involves four main tasks: Characterizing basic virtual objects and development method Developing ontology based simulation methodology for VWE Implementing virtual worlds by case studies Eval uating ontology based simulation methodology for VWE Statement of Problem T he virtual world environment (VWE) is an emerging technology that provides Internet based social networking service s to diverse virtual communities within realistic 3D virtual worl ds of buildings, facilities and environmental areas such as forests ( Kelton,2008; Bishop, 2009; Beck, 2009). It has attracted much attention because of its capabilit y of reducing restrictions of space and time and the possibilit y for building unlimited vir tual world with virtual objects T here are many successful applications of virtual world technology in the areas of medicine (Ann Myers Medical Center, 2010;
16 Rogers, 2009; Virtual Neurological Education Centre, 2007) business (Honey et al., 2009), and edu cation (Campbell, 2009; Oloruntegbe and Alam, 2010). But, it is just beginning to gain interest in the agriculture and natural resource domain (Prada et al., 2010; NOAA, 2007; Schott, 2009; Beck, 2009; Jung et al., 2010). Various challenges exist for devel opment of virtual education and simulation environment, and these include: L imitations of VWE as an educational tool L ack of methods of VWE for agriculture and natural resources N e ed for a more efficient and cost effective approach for building virtual ob jects for VWEs Limitations of VWE as an E ducational Tool Virtual world is an emerging technology used by an increasing number of educational institutions around the world and is considered a technology, environment, medium, learning and teaching tool. The technology has been reported by over 100 institutes for education constructed and used a virtual campus in Second Life ( http://www.secondlife.com ) (Hudson, 2007). For educational purpose, virtual world extends th e classroom into a augmented materials and to accelerate the learning processes through direct emersion within a problem space This contrasts with classic textbooks or multimedia web based e learning prog ram s that teach about the subject but do not provide direct experience within the domain of study. Educational applications in a virtual world enable animation, simulations and role playing, distance learning, collaborative experiences, and experiential pr ojects. The virtual world concepts are used in various domains, including architecture and building, art and design, business, cultural studies, health, tourism,
17 social science, emergency services, and scientific simulations (Swanson and Kane, 2007). Ther e are research studies reporting the experience of student learning and teaching by using virtual worlds. Hudson (2007) taught a journalism class for a group of 20 students in a virtual world. Various discussions which were conducted in the virtual worlds discussion. It was determined that the virtual world engages the students in learning from the start and encourages collaboration and cooperative communication between students. Si milarly Lemmon and Kelly (2009) utilized the virtual world for training students in interviewing skills and it was suggested that interview skills can be enhanced through virtual training in role playing scenarios. In those experiments, although a virtua l class room in a virtual world contributed to learning, communication and collaboration among students, simple mimicking the teaching procedures in the virtual world was insufficient for student to utilize rich virtual objects in their learning Furtherm ore, Senges and Alier (2009) determined that virtual worlds support ed opportunities for education purposes, but it would be a challenge build ing the informal and holistic learning scenario for most education institutions. Gregory and Tynan (2009) provided survey results of 11 students on learning experiences in the virtual world and indicated that students felt that their learning in the virtual world was informative, engaging and educative when the virtual world provided concrete learning scenarios Simultaneously, i n most cases of virtual experiments, various virtual objects are required and implemented for educational purposes. Honey et al. (2009) studied learning experiences in simulating patient care scenarios, and it was reported that a
18 virtual world helped student s learn skills to care for patients But, it required intensive work to building the virtual clinic and props. Similar results are reported by various researchers in education including a constructivist learning environment (G l et al., 2008), a case study of database administration laboratory work (Cranitch and Rees, 2009), the evaluation of a n emergency medical treatment computer game to support triage training for decision making that prioriti zes mass casualties in terms of treatment (Jarvis and Freitas, 2009), science teaching and learning (Oloruntegbe and Alam, 2010), and the experiential learning study (Wood, 2009). Virtual objects for these experiments are critical resources of a virtual world that consume expensive work. There is a problem with sharing the virtual objects in other experiments. There is research reporting a bright but cautious future for learning in the virtual world. A s urvey result from educators in Second Life (Swanson and Kane, 2007) indicates that the virtual world offers great potential for the future Web, but it will take more time to be come a mainstream of learning because users feel uneasy about virtual objects. Wood (2009) suggested that a method of learning by virtual world would need a more cautious app roach because he claimed that the technical problems of the virtual world contributed to student dissatisfaction with the learning experie n ce. Campbell (2009) utilized a virtual world for learning simulations by role playing to improve professional decis ion making among pre graduate teachers taking a unit in professional ethics. It is reported that the virtual world enhanced interest and experience of students in learning. But, he claimed that the effectiveness of the virtual world is limited by lack of r eality of virtual objects, efficiency of collaboration tools, and information for usage of the complex system.
19 Freitas et al. (2009) proposed an evaluation methodology for supporting evaluation of specified learning activities in a virtual world instead o f using a general survey questionnaire. By applying it, he concluded that students were positive about using the tools for supporting collaboration based on shared interest. And, it is claimed that appropriate technical support and well structured sessions in virtual world are essential for providing enriched experiences from the virtual world. Lack of Methods of VWE for Agricultur e and Natural Resources While virtual world technology has been used for learning and teaching in various domains, in the domai ns of agriculture and natural resources it just beginning to gain interest. In the early applications of virtual reality technique which focused on 3D simulated virtual environment rather than social networking of virtual world, Orland et al. (2001) propos ed communication capability for landscape planning decision support systems. G eographic information system (GIS) tools are utilized to represent geographic data such as soil profiles in 3D, and the system facilitated hypermedia and Internet based communica tion. However it is a closed system in which only prebuilt virtual objects are utilized and provides a limited access to the system for highly trained users, and collaboration and communication between users in the virtual environment are restricted. Wi th recent advances of Web technology like the new paradigm of Web 2.0 utilizing dynamic web contents such as blog, wikis, video sharing to interact and collaborate with users (Anderson, 2007 ; Boulos and Wheeler, 2007 ), in the agriculture and natural resour ce domain web applications supporting virtual communities are increasing ly used for educational purposes. An agriculture extension program at Ohio State University utilizes a blog site for gathering agriculture news with a real time
20 repository of informati on by two way communication (Kleinschmidt, 2009). The FAO (Food and Agriculture Organization) site services various audios for international and regional communities through national broadcasters. And, Bemis Schurtz and Bussmann (2009) proposed an e learni ng site using Web2.0 applications for learning about ecosystem s The ecosystem is a real time conversation based system, and provides desktop widget s and web widget s to find an interesting community and to suggest an idea. And, Twitter and mobile communica tion is utilized to share the ideas and improve communications. Virtual world technology facilitate s a Web 2.0 service hosting other application s and services by using visualization. In the 3D web, it provides more intuitive ways to navigate multimedia col lections, and to browse information spaces and document collections in virtual libraries. The National Oceanic and Atmospheric Administration (NOAA) launched a virtual world that facilitated education about the Earth s environment in many ways (NOAA, 2007) In the virtual world users can stand on a plane while going through a hurricane or experience the ice caps melting by standing on them. V irtual islands are utilized to educate people about the Earth's o ceanic and a tmospheric features. In agriculture and natural resources education several virtual worlds are developed as games (Farmville, 2009;SimFarm,1993) within which students follow rules and steps that are predefined to achieve educational knowledge. Prada et al. (2010) proposed a game implemented in the on line virtual world platform, the OpenSimulator, for learning the potential effects of agriculture in the environment such as the impact of fertilization in water quality and production of food. They assessed the quality of the
21 game and user's expe rience by using a questionnaire on challenge, difficulty, control, clearness, improvement knowledge, and awareness of the impact of agriculture in the environment. It showed good acceptance by the users and positive result on enhancement of awareness of im pact in agriculture Ecosystems and forests in virtual worlds are usually represented as a landscape. In virtual world environment like Second Life there are various examples ( http://secondlife.com/ destinations/nature ), which provide diverse forms of nature and environments. They are useful for experienc ing virtual nature and enjoy ing artificial scenes, but the reality of features is limited by graphic resources and capability of the VWE. Lack of ed ucational tools and entities restrict use of them in teaching and learning. Beck (2009) suggested prototypes of virtual laboratories and entities by using the Lyra VWE for agriculture education where students can interact with 3D, simulation driven learni ng experiences modeled after real laboratories and entities. Examples are provided of a virtual greenhouse, soil profile and trees proposing how to design and use them in the domain of agriculture and natural resources Appropriate Approach for Building Vi rtual Objects for VWE s Within the traditional approaches of developing virtual world s by manual programming and hand constructing art works, constructing virtual worlds requires much effort to create structured data of graphic model s and implement complica ted behaviors, relationships and processes among objects. VWEs are developed with different programming languages and different graphic engines, and supporting graphic model formats are various. For example, OpenWondeland (2010, www.openwonderland.org) VW E developed in Java supports model modules complying with jMonkey ( http://jmonkeyengine.org ) a graphic engine,
22 API, and graphic model formats in X3D ( http://web3d.org/x3d ) and COLLADA ( http://collada.org ) are allowed in the VWE. Meanwhile, OpenSimulator VWE is developed with C++, and supports models created with graphic primitives oriented from Second Life VWE server. A model module is a prog ram set indicating the associated graphic model object and manipulation processes for the model. Although it is a useful and powerful structure to integrate graphic models and behavior, it is limit ed in sharing with other VWEs adopting different system for mats. The complicated information in a module targeted for proprietary VWEs can be constructed from data structure s containing abstracted information by using an ontology. Ontology is a collection of concepts and relationships among these concepts in a spe cific domain (Noy et al., 2000), and a promising technology for knowledge reuse and knowledge sharing (Beck, 2003). There are some efforts to apply the ontology based approach to the VWE. Anthopoulos et al. (2010) proposed a virtual learning environment fo r training in profiling, negotiation and crisis management. The environment utilized a rich knowledge base of professional experiences on negotiation methods, crisis management, decision making and legal affairs. An o ntology was adapted to express terminol ogy and instance information related with problem solving procedures and collaboration. Meanwhile, Ahmed and Gracanin (2010) addressed the interaction interoperability issues of 2D input devices for virtual world, and proposed a framework adopting ontologi es of input devices and tasks for abstracting the input and output of various devices and interaction techniques to various tasks.
23 Ontologies are utilized for enhancing interoperability of data and model. Suman et al. (2010) designed and developed a virtu al dental school by using open source virtual world technology (OpenWonderland). It is implemented in the distributed client architecture for maintaining applications of the virtual world flexibly. And, the application architecture was improved by represen ting configurations of virtual objects in XML explicitly that includes the information of model behavior, storage location, orientation and scale. Jeschke et al. (2009) proposed a software infrastructure by utilizing the OpenWonderland (2010) VWE for enabl ing the interlinking and integration of experimental superstructures and simulations of nanotechnology in natural science and engineering. A subject ontology is developed to distinguish experiments from simulations by describing the input and output data o f a simulation and evaluation algorithms explicitly. Sanchez Ruiz et al. (2008) assessed the data interoperability problem in the virtual worlds for high resolution weather data from various agencies, and suggested to use ontologies to manipulate the seman tics of visual representations and operations associated with the data. It facilitated recording data operation by users through their avatars in the virtual world, and supported transform ing them into various resources such as multi language source code, visualizations, and web documents by external plug ins. However, these efforts contributed to utilize the ontology for representing internal system knowledge, not to expand the approach to interoperate the model within other VWEs. For generating and manip ulating graphic objects in virtual world automatically Beck (2009) introduced an approach of ontology based data management and provided prototypes of greenhouse, soil profile and trees. Applying an ontology driven approach
24 to build a virtual world can pr ovide ways to describe the properties and structures of a n object explicitly with reduced efforts, and it can enable to create them automatically. For example, once the prebuilt virtual objects such as soil and plant are stored into the ontology, they can be generated automatically by using the virtual object ontology instead of creating a nother virtual world and objects manually. For simulated models in virtual worlds, there is a need for supporting communication between virtual objects by using embedded models or processes during simulation or interaction in the VWE. Ontology based approaches to developing dynamic models, which explicitly represent knowledge of concepts in system and relations between them, have shown successful results in building simula tion systems generated automatically from a structured model ontology, and authoring tools are used to build and maintain dynamic models without programming (Beck et al, 2008; Kwon et al, 2010a). The ontology based simulation approach consists of a set o f procedures: 1) creating an object relationship diagram, 2) defining concept with symbol, definition, and related information 3) developing a model or relationship between conc epts in the form of mathematical re presentati on and 4) automatically generati ng code for running the simulation. In a similar way, it is possible to develop an ontology based simulation for a virtual object in a virtual world. For eLearning system s such as a forest learning system, interactions between instrument s and nature dri ven objects can be described. For example, a logger tape object used to measure distance from instrument to target object can have a process providing a distance value by interacting with object. The process can be represented
25 as a mathematical expression of vectors of the object's positions. In addition, object behavioral process es such as showing changed tape line when the instrument is moved will be useful for describ ing interactions within and between objects It is possible by abstracting virtual objec t s according to the role of model, view and control. Specific Objectives The objectives of the virtual world simulations developed in this dissertation was constructed to address the following objectives. Implementing virtual worlds by case studies on crea ting virtual forest world with the generated models and the simulated forest population. Evaluating the VWE by conducting a survey on the Virtual L earning Forest and measuring educational achievement Characterizing basic virtual objects which belong to th e agriculture and natural resource domain for educational exper iences and simulation, and developing procedures from design ing to implementing objects Developing an ontology based simulation methodology for VWE by implementing information structure of sim ulation and experiment s with an ontology and using them to automatically generate digital models and build virtual objects across different VWE platforms Approach A Virtual world environment (VWE) is a genre of online community taking the form of a comput er based environment where users can interact with others to use and create virtual 3D objects (Bishop, 2009). Use of the interactive 3D virtual environment and avatar based social networking has increased the number of people using virtual worlds for dive rse purposes including education (Bray and Konsynski, 2007). In particular, it enables creation of a virtual world which is able to deliver educational experiences around the Internet to different users from different countries to interact, provide, gather and evaluate information.
26 The VLF for the forestry laboratory exercise in the virtual world were initially built based on the OpenWonderland platform The VLF is a composite module system composed of the longleaf pine module and terrain model. The level o f detail of the tree module is varied by switching different detailed models according to the distance from the avatar, and resulted in significant increasing the performance of the virtual world system. In the near distance a fully developed 3D model is u sed to enhance the reality of the world. For the mensuration exercise, three field instruments logger tape, diameter tape and clinometer are developed as HUD components which help students to feel like they are using real instrument in the virtual world. A prototype VLF is developed in the traditional way of implementation by manual coding following the proprietary model protocol of the VWE, which gives a solid, concrete virtual world. Three types of virtual objects are developed, which are 1) terrain ob ject, 2) forest object and 3) instrument object. A virtual object of terrain is an object focusing on a graphic model and its geological information, while a forest object is a combined model of graphic model of a tree and dynamic model simulating populati on of trees by generating geological and physical information for each tree based on a single prototype tree model. Virtual objects of instruments are objects which stress their behaviors including interactions with the controller (avatar) and communicatio ns with target objects as well as graphic model and interface for representing data. With the traditional methods, virtual objects are expressed in a proprietary format oriented in the VWE, including information about relationships between graphic model s, behaviors and processes. But, this results in limit ed accessibility to inflexible structure s and information when it is required to create a new virtual world and to modify
27 processes in a virtual object. Behaviors of virtual objects are implemented by u sing pr o prietary languages required by the specific VWE, so it is difficult to understand and share them efficiently. Processes (logic) of calculation s within virtual objects such as instruments can be represented in an independent form as mathematical exp ression s To solve some of the problems that come from the traditional development method s the approach presented here improve s the flexibility of the model building process by introducing ontology based data management technique. Model interoperability i s enhanced in two ways; 1) ontology based simulation model development, and 2) ontology based exchangeable digital content The ontology based data management system is developed to organize information o n domain specific educational system s such as forest field size, tree distribution and population information. I t cooperate s with a simulation mo del generation engine to build virtual world s The ontology based exchangeable digital content system is developed to organize information related with digital con tent such as shape, material and position by excluding proprietary software and data formats. The digital content ontology is created to define digital elements. It contains concepts for extended model feature such as model behavior and related processes t hat describes model interaction with the external environment. The VLF system has been evaluated by different user group through testing different aspects of the system during all system development stages. Evaluation tests consist of analytical evaluation by an expert group and usability evaluation by a student group. At an early stage of development, a prototype model was evaluated by experts familiar with forestry field experiments and teaching. This evaluation was used to
28 resolve questions about the bas ic approach and to improve 3D object design and interaction design. During the analytical evaluation, virtual object format and design were discussed and selected from diverse possible graphic formats. Designs of interaction and user interface were issues discussed over the entire development stage, and they have been improved by modifying systems based on results of discussions. With the VFL system, a field test for achieving qualitative evalu a tion of the system performance has been conducted with 20 compu ters having the same system specification and running under same network environment. The test focused on estimating system performance and system speed and providing system specification guideline Usability testing for system evaluation and evaluation of lessons learned are conducted with students taking a related class The questions of system speed, level of intuitiveness and usefulness are included on the evaluation items, and it helps to get quantitative scores evaluating the value of the system for s tudents Dissertation Layout Chapter 2 describes the development of the Virtual Learning Forest ( O bjective 1 ) and illustrates virtual objects developed for experiment s and main features of the system. Evaluation studies (Objective 2) are presented to expl ain the usability of the VLF system developed by the proposed method. Chapter 3 describes approach es for abstracting and characterizing virtual object s to identify the role of model elements ( O bjective 3 ). Chapter 3 also explores several problems with trad itional development method s for virtual object s Chapter 4 describes an ontology based approach to improve model portability and interoperability ( O bjective 4 ), including example s of defining feature s and function,
29 creating digital content, developing a m odel ontology, and generating a virtual object automatically from the content in the ontology Chapter 5 summarizes conclusions and contributions, and identifies future directions.
30 CHAPTER 2 VIRTUAL LEARNING FOREST: AN APPROACH FOR DEVELOPING EDUCATIONA L EXPERIMENT S AND SIMULATION S WITHIN A VIRTUAL WORLD ENVIRONMENT Background Virtual World Environment (VWE) as an eLearning tool attracts much interest with its capability for online based communication and collaboration and possibility for mimicking a rea l world with virtual objects (Callaghan et al., 2008; Hodge and Collins, 2010) However, the high level technical knowledge required and limited resources available for constructing virtual object s becomes a challenge for utilizing the technology wide ly T he Virtual Learning Forest (VLF) project is an effort to adopt the VWE for forestry education in order to improv e traditional computer based training. It was designed to provide various virtual laboratory experiences to student without limitations of space and accessibility and provide opportunit ies for improving student field skills. In the VLF, students create a n avatar that is a virtual person object representing themselves in the virtual world. Then s tudents take measurements in order to estimate the tree height and tree diameter of the forest. Students begin by identifying a sample plot a small region within which all the trees are measured. Next students utilize virtual tool s to measure tree diameter To measure the tree height with virtual clinome ter, students measure a distance from the tree at first, and then use a clinometer. The measurements are recorded in the logger note. T he VLF for the forestry laboratory exercise was developed with in the OpenWonderland platform. It is a composite module s ystem consisting of environment modules and field instrument based on graphic model s and Java program code. Forest and terrain module s are two environment modules, and a forest of 400 longleaf pines is
31 created that occupies a 2 acre terrain. Field equipmen t is used to measure tree height and diameter, and includes a logger tape, diameter tape, and a clinometer that mimic shapes and behavior of real instruments. A s ampling area tool is used to identify a region of the forest in which all the trees are measur ed and includes a center pole, boundary tape and tree tags. U ser interface objects including a logger note and tool bag facilitates recording data from measurements and provid es help to students for guiding their activities Virtual World Environment (VW E) VWE is a computer based environment where users can interact with others, and use virtual 3D objects (Bishop, 2009). Use of the interactive 3D virtual environment and avatar based social networking has increased the number of people using virtual worlds for diverse purposes including education (Bray and Konsynski, 2007). In particular, it enables creation of a virtual world which is able to deliver educational experiences around the Internet to different users from different countries to interact, provid e, gather and evaluate information (Ramasundaram et al. 2005) There are various VWEs including commercial packages and open source environments. Second Life (Linden Lab, 2003) is a popular and mature environment which has been utilized for virtual campus es of many universities including the U niversity of Florida, Harvard U niversity and Stanford U niversity However, it requires student to subscribe to use the service and provides only an open network based on islands OpenWonderland landscape) Like other commercial package s such as OLIVE (2010), the limitations of cost and accessibility for building virtual worlds for experimental and educational purpose s encourage using open source technologies instead such as OpenSi mulator (OpenSimulator, 2007), Croquet
32 (Croquet, 2007) and OpenWonderland. OpenSimulator is a platform (Microsoft OS) dependent system, so compatibility with other operating systems can be restricted. Croquet is an open source project based the programming language Smalltalk (Smalltalk, 1972). OpenWonderland is an open source tool and environment written in Java for building and running virtual worlds. Thus, it has better cross hardware platform independency compar ed with OpenSim ulator and offers better o pportunities to utilize the wider range of resource s available in Java Asset management architecture in OpenWonderland supports extendibility. Graphic models can be developed with common external graphic tools like Blender (Blender, 1995) and imported in to OpenWonderland utilizing formats such as X3D (2004) and COLLADA (2004) which are standard format s for sharing graphic models through the World Wide Web (WWW). Behaviors of graphic models also can be implemented by using Java Script internally Virtual Learning Forest (VLF) World For forestry education, the VWE shows much potential for offering immersive learning experiences that improve on traditional computer based training. It can offer laboratory experiences that complement real laboratories which m ay be expensive, dangerous, or inaccessible to students. The Virtual Learning Forest (VLF) project (Bannister, 2009) attempted to create a new forestry educational delivery paradigm by adopting the VWE technology. It was hoped to overcome limitations of co st, time, distance and space. T he project aimed to achieve the following objectives (Bannister, 2009) : (1) "s trengthen and expand the skills students learn in their existing field exercises ", (2) "o ffer ing a variety of virtual laboratory experiences to a large number of students across the US and internationally so that they can build experience in forest
33 types they would not have the opportunity to physically visit ", (3) "p rovid ing students opportunities to interact with their peers at other institutions via collaborative exercises early in their training, the importance and advantages of developing relationships with distant colleagues in natural resource management and (4) "o ffer ing these thereby expanding learning opportunities for a large number of students at little cost ". During the beginning stage of system development ther e were discussions about the effects of educational environment changes on the new paradigm as well as technical issues. And, they contributed to expand consideration s for roles of the system in the educational environment and for the required features of system An example describing a conception of role s of the VLF and its possible relationship s with educational environment is given at Fig. 2 1 Changes in educational policy or situation stimulate the development of the Virtual L earning F orest system in the school domain. The system consists of a server, virtual forest libraries, model ontologies related with instrument, tree growth or forest generation, and GIS data for terrain and forest. The s chool domain includes teacher, student, and system manager a nd researcher and distance education student s have close relationship with the school. Other schools a nd international institutes utilize the VLF for education and cooperate to enhance the learning program and to improve asset s of the virtual environment The s chool or system manager respond s to the requirements of education methodologies according to the changes of educational situation, policy, and technical improvement. And, requirements from community affect to decide the direction and
34 content of educ ation in school. Teacher provides educational material to the learning system and set plans for field experiments. Student s use the learning system for learning and developing field skill through the virtual environment. With responses from student and fee dback from other users, researcher s evaluate the virtual learning system and learning contents. Results of the evaluation are used for improving learning system and enhancing the learning environment. As future contributions of the VLF, s tudents graduated from school seek chances to use knowledge and field skills at work places by being employed. Distance students who work at the job and learning field skills through virtual learning system contribute to communities by using learned knowledge. A n e xample of a conceptual model diagram is shown in Fig. 2 2 concentrating on the functionalities of the VFL. A big circle represents a system boundary of the VLF, while small circles in the big circle indicate functions of the VLF. There are resources of information and data at the left side of system boundary. The f orest generation process consists of terrain creation and tree creation. These processes utilize topological and environment data from the GIS. Data created for the forest and obtained from the instrument s are associated with data storage. A prototype virtual world consists of a forest ecosystem and virtual educational instruments for measuring and recording information about tree s in the forest to support real field experiment activity in classes. Objects in the domain are categorized into natural resources (tree and terrain) and data gathering instruments (clinometer, logger tape). A tree is an essential virtual object for represent ing a rich ideal forest, and a forest can be enhanced by combining appropri ate generation models. A terrain model can be enhanced by implementing and adopting a generation interface using a land elevation model based on GIS data
35 Figure 2 3 illustrates the prototype VLF client/server architecture. Server side components of OpenWo nderland consist of a game server (DarkStar) on the web server container and a voice communication server (jVoice Bridge). Client components consist of system layers (game server protocol, jMonkey 3D graphic engine, and VWE client) and virtual world laye rs. System layers enable communication with server components, to render virtual objects, and to maintain system assets in the local (student) computer. Virtual world layers implement the VLF world object and related applications which are designed to be independent to the world itself. Application data are provided through server/client communication from the server. Main Features of VLF The VLF is a composite module system including scene components instrument components and user interface components Scene components are used to build the virtual forest, which includes the longleaf pine tree module, the terrain model and the forest module by combining the two components in the scene. Instrument components including a logger tape, a clinometers and a d iameter tape, are developed as a head up display (HUD) component User interface components are 2D virtual objects formed in HUD object without associating with a 3D graphic model, and include the Welcome Board, Tool Bag, User Guide, and Logger Note object Scene Components for Virtual World The forest world consists of three scene components that are structured object s in the scene; 1) tree object, 2) terrain object and 3) forest object. Tree o bject Detail level and style of tree model affect the performan ce of virtual world. Highly detailed (high resolution) tree models improve the reality of scene in the virtual world
36 (Palubicki et al 2009), but require considerably complicate d features and delicate texture resources which can overburden graphics hardwa re resource used for rendering the required detail. Thus, it is critical to choose a balance between reality and performance for the target virtual world. A longleaf pine tree model from the SpeedTree (SpeedTree, 2002), a fully developed commercial 3D tre e library, was modified and adapted by reallocating shapes and adjusting resolution of textures to adjust the quality level of the model. As a leaf of the longleaf pine tree is a needle shape, it is usually featured as a group of leaves rather than each in dividual leaf shape to enhance the graphic performance but keep a high level of visual quality For better system performance, it is helpful to use various shaped and styles of a 3D model, which can range in level of detail from a low resolution billboard 2D plane to a fully developed high resolution 3D model in which every branch and leaf is represented by a vector object. The level of detail (LOD) technique, which dynamically manages the appearance of detailed features, is applied to the tree model. One form of LOD is the continuous LOD (Lindstrom et al, 1996) which controls the number of vertices When it is possible to use the technique, it is created automatically in the gami ng engine used to render the VWE. The discrete LOD technique switches different models at various levels of detail for the same object and can significantly increase graphic performance, even more than the continuous LOD. It is applied to the longleaf pin e tree module with four different style models, which include three 2D plane tree models and one partial 3D vector
37 model. The 2D models use an image of a longleaf pine as a texture and are created by intersecting image planes having different angles of ori entation. They are set to appear at the range of 0m, 30m, 60m and 90m. In range above 90m a one plane bitmap tree model is used. In the near range within 0 30m from the avatar position, the highest resolution 3D model (617 edges, 83 faces, 15 materials) ap pears with a cone shaped trunk textured with fully detailed bark. Terrain o bject A terrain model is built in a plane object representing a 2 acre area. The model represents only the surface of terrain, and is mapped with a simple grass texture. The 2 acre area of the terrain is chosen as a small but realistic forest size, while considering the system performance of the virtual world containing 400 trees mapped onto the terrain. To improve the user experience, the topographical characteristics of longleaf p ine habitats are considered for build the terrain model, and the topology involves area elevation and steepness. Th is information was collected based on the typical longleaf pine habitat. Most longleaf pine habitats occur around the southeast ern United St ates and m uch of the state of Florida state consists of longleaf pine habitat (Peet and Allard, 1993) The elevation data set for Florida state is created by using the digital elevation model (DEM) of Florida that is published with 30m resolution by the U.S. Geological Survey (USGS DEM, 1999). Using a geographical information system (GIS), the DEM dataset is overlapped with a Florida county boundary dataset (USCB, 1990) and a Florida forest inventory dataset created by the USDA Forest Service (USDA, 2009) which contains more than 20 data element s including tree sampling results such as
38 height and volume of a selected tree. By processing raster analysis using GIS, mean and variation of elevation in the longleaf pine habitat s are obtained. Based on those v alues, some referencing vert ices are generated by using a normal distribution and some manual changes are added to fix odd elevations. They are added to the elevation data set of the terrain plane model, and it is refined into a terrain model using groups of triangulated planes. Forest o bject A forest module constructs the virtual forest world by compositing the longleaf pine modules and the terrain model in the scene of the virtual world. It generates and loads 400 tree models and a 2 acre terrain model, and applies scale factor and transformation factors to 400 trees. Tree distribution and scale information are created and utilized within the module. All 400 trees are created from the same original prototype tree model, but each tree is scaled to a parti cular height and diameter. A tree distribution is created with random numbers. Two random number s sets for 400 trees are generated by the Java random number generator ( http://do cs.oracle.com/javase/7/docs/api/java/util/Random.html ), and the real numbers between 0 and 1 represent coordinates of trees in the 1M x 1M scale terrain. Actual coordinate values are defined by multiplying actual scale of target terrain (90M x 90M). Avera ge tree age in an area can be obtained by GIS data (Florida forest inventory data from USDA Forest Service) analysis with longleaf pine habitats data. Sizes by tree age and scale variation of tree are generated by applying a normal distribution to the stan dard age height relation (Jack et al., 2005) and the height diameter relation (Platt et al., 1988).
39 Instruments for Mensuration Mensuration involves measuring physical parameters (e.g. height, diameter) of tree s in a forest. During the field experiment fo r tree measuring, mensuration is conducted with various instruments in accordance with the measuring procedures. The l ogger tape, clinometer and diameter tape are common instruments for mensuration which are utilized for determining the quality and dimensi ons of trees. The s ampling area tool is a combined instrument consisting of sampling boundary tape, center pole, and tree tag. Most instrument objects consist of a graphic model of the instrument and student interface HUD object displaying enlarged measure ment scale to make it possible for students to read the instrument scales The graphic model of an instrument is presented in the virtual world, so other students can see the appearance of model. Each student own s a set of virtual instrument s HUD object s are only shown at a student's individual A Head up display technique is utilized in the VWE to develop the instruments, and shows a transparent display that presents data without requiring the student to look away from the normal viewpoint. Students can view information by looking head up and forward instead of having to angle down to look at lower instruments because information is displayed in the direction in which the avatar is looking. Samplin g area tool A sampling area tool (Fig. 2 4) is utilized to represent a current experiment area where students undertake measurement. All trees within the sampling are to be measured by the student. Measurement activities with the instrumen t are conducted within a small circular plot area having around 40 ft radius over the whole experiment
40 area The radius can be changed by students if necessary The sample are tool consists of a center pole, tree tag s and area boundary tape. A center pol e is a graphic model representing a center of the sampling area. The location of the center pole is set by the student by clicking a certain ground point. The location of this point can be picked arbitrarily by the student. The s ampling area boundary is m arked by a boundary tape that appears on the ground at 40 ft away from the center pole. The tape is composed of line segments that form into a circle A t ree tag is a graphic object containing a unique identification number for each tree, and is placed bes ides each tree within the sampling area. Students use the Tree tag ID number to record measurement data for each tree. Logger tape In the real world, a logger tape is usually a 100 foot long metallic tape that is spring loaded for retraction into a metal box (much like a home measuring tape) which has distances marked in feet. The end of the tape is often fitted with a nail, so that it may be fastened into the bark of a tree, enabling a single person to measure the distance from a tree to a point on the ground. The logger tape is implemented in the VWE with a tape graphic model and a HUD object (Fig. 2 5 ) The tape graphic model appears at the scene when a student activates the tool. Tape line from tape body to a target object is represented with a tempo rary line object, and is appears as a straight line (sagging effects were not incorporated) A n user interface for reading a measurement from the tape is a HUD object that shows the enlarged view of the instrument including a part of tape body and tape lin e indicating a measured value. By using an image of the instrument, i t helps students feel like they are using a real instrument in the virtual world. It shows the
41 distance measurement in feet and meter scales on left and right side at the head of the tape as the tape is moved forward and backward from the target tree. Clinometer In forestry, a clinometer is used as an angle measuring device used to determine the height of a tree. It uses simple trigonometric relationships to determine tree height given t he distance the student is standing from the base of the tree (this is determined by the logger tape) and the angle from the ground to the top of the tree. This angle is determined by viewing the top of the tree through a viewfinder available in the clino meter. The cli n ometer typically has two scales: a percent scale for the tangent value of the angle (on the left) and a 1:66 scale (on the right). The 1:66 scales gives the direct height reading at location 66 feet away from the tree. The clinometer is rep resented in the virtual world with a clinometer graphic model and a user interface for reading measurement s (Fig. 2 6 ) The graphic model appears in front of the avatar when the instrument is activated. The user interface is formed as a HUD object displayi ng a dynamic measurement of both scales. Scale rulers in the object are represented with line graphic elements, and changes of view angle are reflected instantly by updating location of lines and referenced values of the ruler. Diameter tape A diameter tape is a tape measure used to determine the circumference of the trunk of a tree. A diameter tape has a smaller and more compact shape compared to the logger tape, and gives measurements in inches and centimeters when wrapped around the tree. By conventio nal standard, m easurement of the tree diameter occurs at and it is specifically at 4.5 ft above the ground.
42 In the virtual world, the diameter tape (Fig. 2 7 ) is created with a tape graphic model and a user interface HUD object. The tape m odel appears in front of a tree where student clicks on the tree at the point where the measurement is to be taken When clicking onto the tree, student can refer the height of the point represented at a tape ruler in the HUD object. The HUD object contai ns two rulers for diameter measurement and height measurement. The height measurement is represented in a vertical dynamic ruler. The diameter ruler is mimicked by displaying a real image representing a tape wrapped around a tree, and the tape ruler and n umbers are updated as the measurement changes The reading value is calculated internally by accumulating distance s between neighboring vert ices on an imaginary mesh plane which intersects at the trunk of tree. User Interface Objects Several user interface objects were developed to provide support to students while conducting experiments. These include the Welcome Board showing objectives for learning Instruments Guide to explain the tool usage a Tool bag for selecting instruments, and a Logger Note tool to record measurement data. They are implemented in HUD objects, and no 3D graphic model is associated with them. Welcome b oard A welcome board object (Fig. 2 8) is a user interface showing information for user s when they first access the VLF. It is imple mented as a HUD object which contains the owner's symbol image, title of the virtual world (VLF), and version of the VLF, and objectives which the student s are to undertake during an experiment. It appears as a dialog in the middle of the OpenWonderland ap plication window at the begin of each session
43 Tool b ag A tool bag object (Fig. 2 9) is a HUD object which appears at the lower right corner of the application window It is an image panel showing a bag of tools which are the instruments which can be util ized in the VLF. It contains the logger tape, diameter tape, clinometer, GPS (which matches with the locator object providing current location of the user in the virtual world), and logger note. An instrument is activated/deactivated by mouse click actions and a snapshot image is shown when the mouse pointer is placed over an instrument image. The snapshot helps the user to remember how the instrument is utilized in the VLF. Instrument s g uide A n instrument guide object (Fig. 2 10) is a sub tool of the tool bag object, and it appears as an image panel describing step by step procedure for us e of the selected instrument. It consists of snapshots of all phases of the tool use and a student can learns detail s about control s for using the instrument Logger n ot e A logger note object (Fig. 2 11) is a virtual object which records student's data entry It is developed as a HUD object containing data fields and data table. It is utilized for keeping the recorded measurements during the experiments. And, the data can be sent to the remote database or saved in a CSV (Comma Separated Values) file. It is designed to let students record their tree measurements for all trees in a particular sampling area. To track the sampling area location and trees in the boundary, plot number and plot center coordinate in the virtual world can be recorded in the log g er note ). A d ata table contains columns for tree measurements recorded by tree ID
44 number, DBH (inches), distance (feet), two upside/downside clinometer readings for tree hei ght measurements. Evaluation of VLF System The developed VLF system has been evaluated by different user group s by testing different aspects of the system during several development stages. Evaluation tests consist of qualitative evaluation by an expert group and usability evaluation by a student group. The evaluations were done on a fully implemented version of VLF running under OpenWonderland. Evaluation of System during Development At an early stage of development, a prototype model was evaluated by e xperts that are familiar with forestry field experiments and teaching. This evaluation was used to resolve issues on the basic approach and to improve 3D object and interaction design. During the qualitative evaluation, virtual object format and design we re discussed and selected from diverse possible graphic formats. Designs of interaction and user interface were issues discussed over the entire development stage, and they have been improved by modifying the system based on results of discussions. F igure 2 12 show s an example of the modified diameter tape tool. The tool was designed with a grid plane and a HUD interface to display the position of the measured point on the trunk (Fig. 2 12 a). With discussion results on the user interface design, the grid plane was replaced with tape around the tree trunk. And, a new tape for showing the height of measuring position was added to the HUD interface to improve interaction efficiency for users (Fig.2 12 b).
45 Evaluation of System Performance With a prototype of the VFL system, the system performance was continuously tested with diverse computer systems, because the VLF took much computer resources to initialize the virtual world with 400 detailed trees requiring heavy, instant memory usage of 370 MB (32 bit syste m) ~ 560 MB (64 bit system). The test s focused on estimating system performance and system speed and providing system specification guideline. T able 2 1 show s the diverse computer specifications and the system loading speed results tested during Jul ~ Au g, 2011. In the table, 8 resource specifications of computer system are presented with information o n machine name, operating system, CPU, RAM size, and graphic card. Two types of personal computer s desktop s and laptop s are utilized for testing, which ar e connected on wire d or wireless I nternet connection s The G3D rating explaining the 3D graphic ability of a graphic card is referred from a video card benchmark site ( http://videocardbenchmark.net ) as of Jun, 2013. The VLF launching speed is the measured time to complet ely to display every virtual object in the world. The launching times ( except case 4 ) represents an average time over 20 times of trials at each machine that is utilized at the development phas e. With the machines of case 4, a field test for achieving qualitative evalu a tion of the system performance was conducted with 20 computers having the same system specification and running under the same network environment at a computer laboratory located in the University of Florida A griculture and B iological Engineering D epartment The test resulted in 22 seconds of average system loading time. V irtual instruments operated the way they were intended, with satisfactory performance speed, and all users cou ld shared and collaborate in the experiments from their individual workstations
46 T hrough tests on diverse computers, the results showed that higher speed of CPU and graphic card contributed to improve the system loading speed as did bigger RAM size. Being a server client system type system it was much affected by I nternet connection speed at the first loading of virtual components. High specification graphic card s howed better system performance, but it also needed good support of fast CPU s and large RAM size. Evaluation of System Usability : System Design Usability testing and evaluation of lessons learned w ere conducted with 30 students studying forestry and agriculture at the U niversity of Florida. A survey was used that includes questions about system speed, level of intuitiveness and usefulness, and it was intended to get quantitative scores measurin g the value of the system. The evaluation form is presented at A ppendix A. The questionnaire has three categories of questions about; 1) user and computer system (Table 2 2 ), 2) user experience of virtual experiment (Table 2 3 ), and 3) user satisfaction for the system (Table 2 4 ). And, most questions (except about computer system type, level of experience, and simulation time) have a seven point scale forma t where 1 means not good 7 means very good, and 4 indicate s middle of satisfaction value. Category A (Table 2 2) consists of 8 questions about user background and computer system the student was using M ost users have modern computers system manufactur ed within about the past two to three years, and most have a good graphic supporting system (critical for good performance of the VWE graphics) 90% of student s in the survey already had knowledge of field experiment s but they had little prior experience with in 3D game s (1.2/7.0). Al though they have little direct experience
47 with virtual world s (1.9/7.0) more students knew about virtual world system s such as Second Life (2.7/7.0) After taking instruction about the VLF system prior to use of the VLF th ey showed high expectation for positive result after learning the system (5.3/7.0), and expected to see moderate reality (4.1/7.0) from the system. In category B (Table 2 3), there are questions about each experiment step with a virtual instrument Questi ons were designed to get answers about efficiency and usefulness of forest data gathering experiment s in the virtual world. With respects to the user guide and virtual tool box, the tool box design was found to be highly intuitive, but caused problems beca use of its size and position in the screen. Users reported that the user guide gave helpful tips. The measuring area tool received good scores on displaying area center pole, tree number tag and boundary lines but scored lower for realism and intuitivenes s because of the complexity of the tool that is co mposed of three different components. The diameter tape tool obtained good scores on its design and usefulness but also scored lower on intuitiveness. Responses for the logger tape tool show ed high satisfa ction for realism, intuitiveness and usefulness because of its simple design and ease of us e Result s for the clinometer tool show ed similar results to the logger tape For reporting note tool, there are many replies that gave low score on its intuitivenes s, but the score of usefulness was high. Most user s had difficulty using virtual instrument at first by following the user guide because they were not familiar with the tools. As they gained more experience users felt that their proficiency for using the tool increased. This is to be expected, but improved tool design can increase the rate at which students learn how to use the tools.
48 These results also need to be compared with real world tools (for example, compared with how quickly do students learn to use a real clinometer). C ategory C (Table 2 4) consists of 10 questions about overall experience of the virtual experiments. Users reported little difficulty on following instruction in the user guide (3.5/7.0), and item s on ease of controlling virtual to ol s obtain low score (3.2/7.0). Realism of virtual world needs to be improved (3.7/7.0), and students requested that there would be more materials helping user to understand content in the virtual world (3.8/7.0). However, users experienced improvement on their proficiency on using the virtual system. The measurement completion time was reduced from 523 seconds (first trial) to 396 seconds (last trial). They obtained good impression on the implementation of the VLF (4.8/7.0) than their expectation, and w ere satisfied with completeness of the system (4.4/7.0) and result s of the experiment (4.3/7.0). Evaluation of System Usability: Learning Efficiency Learning efficiency of VLF was tested to evaluate how well students learned from the system The test measured how the VLF helped student s to learn field skills, and the results came from two different tests at different experiment locations. Students conducted two experiments in order of sector A and B, and they had enough time between the experiments to gain exp erience with virtual tools and procedures. It is assumed that students improved their skills with iterative experiences. The hypothesis imposed in this evaluation design is that the experiment result of sector B will show better accuracy then the experimen t result of sector A. Evaluation was conducted by comparison between measurements of tree height and diameter at breast height. Two experiment locations (Fig. 2 1 3 ) are selected randomly in the map of the target forest generated in the virtual world. Secto r A is a circular area with 40ft radius
49 which is located at middle of north west area of the forest, and there are 19 trees whose heights var y in range fr o m 15 m ~ 24 m and whose DBHs are in range of 14.5 cm 35.0 cm. Similarly sector B is located at mid dle of south east area of the forest including 16 trees. Tree heights and DBHs are various in range of 21 m 26 m and 31 cm 34 cm. Each student was expected to enter into the virtual forest with given coordinates of center position of each sector A and B. They were instructed to follow below procedures during the experiments. 1. Once logg ed into the forest, conduct experiments in sector A and sector B. 2. Find the location of center position of each sector with the Navigator tool of the VLF, and then create a sampling area with the Sampling Area tool. 3. Measure the DBH with the DBH tool by seeking the height on the tree from ground with the Height Pole tool, and record the value at the Logger Note tool. 4. By using the Logger Tape tool, find a good location and me asure the height of the tree with the Clinometer tool. And then, record the distance from the tree and height of the tree at the Logger Note tool. 5. Complete the measuring and recording for all tree in the sector A. 6. Record the measured data in the Logger Not e tool, and present the data. 7. Compare the experience of using instruments with the user guide for instruments. 8. Repeat the step 2 6 for sector B. There was no given time limits for students to complete the measurements The test focused on evaluating how accurately user s gather data on tree diameter and height using the VLF, and the result came from the comparison of means and standard deviations of known (by the system) tree height and DBH in the two sectors.
50 The test was conducted by 30 students indivi dually using the same generated forest (same trees and same tree sizes) and it was designed that experiment of sector B follows by experiment of sector A. Measuring results are summarized in mean and standard deviation of each tree in two sectors, and th ey are shown in T able 2 5 and 2 6 respectively. Measurements at sector A are shown in Table 2 5. There are actual and measured value of 19 trees in the sector. T ree number is assigned during generation of a forest in order of tree size from small tree to large tree. It indicates that sector A consists of groups of small tree among 400 trees in the forest. Distribution of most tree heights range s from 15 m to 20 m. Generally, tree trunk diameter at breast height is proportional to the size of the tree. M easured values are given in mean and standard deviation for tree height and DBH. Average difference of actual tree height and mean of measured values for trees in sector A is 0.486 m, and standard deviation is 2.033. For DBH, average difference of actual t ree diameter and mean of measured values is 0.677 cm, and standard deviation is 1.95. In Table 2 6, measurements at sector B are shown, and they came from measurement of 16 trees. The tree group consists of bigger trees than average size tree among 400 tre es in the forest. Average difference of actual tree height and mean of measured values for trees in sector A is 0.410 m, and standard deviation is 1.740. For DBH, average difference of actual tree diameter and mean of measured values is 0.523 cm, and stand ard deviation is 1.49. To test the agreement of measurement between actual values and measured values of tree height and DBH, some statistics values indicating the agreement are
51 calculated and compared. The mean squared error (MSE), the root mean squared e rror (RMSE), and the mean absolute error (MAE) represent the error between actual value and measured value. Nash Sutcliffe (1970) model efficiency coefficient and Willmott (1981) agreement index are utilized to describing the accuracy of student measuremen t. The equations are given below. where, : the actual value : the measured value by student : the average of the va lues The MSE and MAE are classical measures of agreement that eliminate the problem of compensation between under and over prediction. The MSE has an advantage to be decomposed into separate contributions of error, so it is useful for indentifying the so urces of error. Usually, it is convenient to work with the RMSE. On the other hand, the MAE is a measure of agreement to avoid compensation between under and over prediction. It is useful to examine overall model error in simple cases. The EF and the INDE X are utilized to assess the predictive power of student measurement. If the student measurement is perfect, then = for each situation i
52 and EF = 1 and INDEX = 1. Thus, the EF or INDEX value that is clo se to 1 indicates that the student measurement is close to the actual value. In Table 2 7 it is shown the comparison results of error and measurement agreement. Errors of student measurement for tree height are reduced from 0.263 (MSE), 0.513 (RMSE) and 0 ,486 (MAE) at sector A to 0.216 (MSE), 0.465 (RMSE) and 0.410 (MAE). It indicated that the errors of MSE, RMSE, and MAE are improved by 17%, 9% and 15%, respectively. The EF was increased from 0.905 to 0.931, and the INDEX was also improved from 0.975 to 0 .983. Both values get closed to 1 that means the perfectness of the measurement. The results show that students are able to improve their measurement accuracy of tree height at sector B with previous experience of sector A. Meanwhile, errors of DBH student measurement indicated much improvement of students' field skill. The errors of MSE (0.334), RMSE (0.465) and MAE (0.523) at sector B are less than the errors of MSE (0.0.556), RMSE (0.746) and MAE (0.677) at sector A. The improvements of errors, MSE, RMS E and MAE, are 39%, 37%, and 22%, respectively. The EF was increased from 0.977 to 0.982, and the INDEX was also improved from 0.994 to 0.996. By comparisons of measurement errors and efficiency, the results showed that students improved their measurement for tree height and DBH with reducing errors. Through two experiments at two different sites, it seems that students obtained enough experience at first measurement trial in sector A, and then the proficiency on the virtual tools contributed to enhance stu dent's performance at the later experiment in sector B.
53 Based on the result of error comparison, it is verified that students improved their virtual field skills with iterative experiences within the VLF. In the future, further evaluation in which stude nts' experiments at real world and virtual world can be compared may explain how the VLF can contribute to improve students' field skill in real world. Summary A virtual learning forest (VLF) was developed by using virtual world technology to enhance edu cation experiences of forest data collection (tree size measurements) without restrictions on time, distance and cost. VLF consists of three scene components tree, terrain, and forest and virtual instruments for mensuration including a diameter tap, logg er tape and clinometer. Scene components are designed by focusing on realism of the natur al environment. D ata o n the topolog y of the longleaf pine ecosystem was utilized to build the terrain and forest model. LOD technique is applied to the tree model for providing rich detail of tree in the virtual world from different distances Virtual instruments mimic field equipment with tool graphic model s and a HUD view for reading measurement. Some evaluations were conducted during the development. The result of expert evaluation for the instrument design helped to enhance user acceptance. The system performance test provided various performance guide s according to CPU, RAM and graphic card type. And, a performance test with 20 computers having same specification s showed that the VLF worked in a stable fashion and with good performance. The system usability evaluations for tool design and us age ga ve a result that the user interfaces of instrument represented reality of the tools well, but more support was needed t o help students understand the tools. And, it showed good
54 acceptance and overall satisfaction for the VLF by students The evaluation for learning efficiency was conducted at two random sites to compar e the improvement of tool use proficiency. The results showed that students reduced the measurement errors at the next site with the experience of learning at the first site.
55 Table 2 1 List of average launching speeds of VLF and computer specifications (tested at July ~ Aug, 2011) No. 1 2 3 4 5 6 7 8 Ma chine Dell Precision 370 Dell Latitude D830 Dell Precision T3400 Dell Precision M6600 Sony Vaio VGN TT16LN Sony Vaio VPC Sony Vaio VGN Z520N Apple Macbook Machine Type Desktop Laptop Laptop Desktop Laptop Laptop Laptop Lap top OS Windows XP Professional (32 Bit) Windows 7 Enterprise (64 Bit) Windows 7 Enterprise (64 Bit) Windows XP Professional (64 Bit) Windows 7 Professional (32 Bit) Windows 7 Professional (32 Bit) Wind ows 7 Ultimate (64 Bit) Mac OS X v10.5 Leopard CPU (G H z) 2.8 2.2 (dual) 3.2 (dual) 2.5 (dual) 1.4 (dual) 1.8 (dual) 2.26 (dual) 2.4 (dual) RAM (GB) 4 4 8 4 2 2 4 2 Graphic Card Radeon X1650 (256MB) NVIDIA Quadro (256MB) NVIDI A Quadro (256MB) Mobile Intel QM67 Express Mobile Intel 4 Series (GS45) Mobile Intel 82845G NVIDIA GeForce 9300M GS Intel GMA X3100 Internet Connection Wired Wired Wired Wired Wireless Wireless Wireless Wireless G3D rat ing (1) (1 8300 ) 112 258 258 65 40 12 85 122 VLF Launching Time (2) (Sec) 2 4 18 15 22 (3) 2 1 19 14 17 (1) Resource from h ttp://www.videocardbenchmark.net (as of 06/10/2013), Higher number means better quality (2) Average spendin g time to launch the VLF for 20 attempts (3) Average spending time to launch the VLF for 20 computers.
56 Table 2 2 List of questions and scores of evaluation form: Question Category A Item Questions Mean Std ( 1 ) U ser's computer resource A1 What is the operating system of your computer ? (Windows, Mac, Linux) A2 How old is your computer? (can assume CPU, RAM, Graphic card year ) 2.7 1.2 (2) U ser's computer ability A3 How long have you played with 3D game? (year) 1.2 0.5 A4 How well do you know about Virtual World? 2.7 1.1 A5 How much have you experienced Virtual Learning Environment? 1.9 0.8 (3) U ser's background A6 Did you take a class course related with field experiment? 0.9 (4) U se r 's expectation A7 How much do you expect t o get positive results on lear n ing after you are instructed about VLF? 5.3 0.8 A8 How much reality do you expect to experience before you start VLF? 4.1 1.2
57 Table 2 3 List of questions and scores of evaluation form: Question Ca tegory B Item Question Mean Std (1) Step #1 : user guide & VLF tool box B1 How do you define the helpfulness of VLF user guide? 4.9 1.1 B2 How do you define the accessibility of VLF Tool Box? 3.8 1.0 B3 How do you define the intuitiveness of VLF Too l Box? 6.1 0.9 (2) Step #2 : measuring area tool B4 How do you define the usefulness of tool instruction at screen? 3.3 0.8 B5 How do you define the correctness of displaying area center and tree number tags? 5.5 0.7 B6 How do you define the clearne ss of area boundary? 6.1 0.5 B7 How do you define the intuitiveness of the tool usage? 2.8 0.5 B8 How do you define the realism of the tool? 2.7 0.6 (3) Step #3 : DBH tool B9 How do you define the intuitiveness of the tool usage? 3.3 0.8 B10 How do you define the realism of the tool design? 5.2 0.4 B11 How do you define the usefulness of interface design for reading measurement? 5.9 0.5 (4) Step #4 : Distance measuring tape tool B12 How do you define the usefulness of tool instruction at scree n? 5.2 0.3 B13 How do you define the intuitiveness of the tool usage? 5.6 0.3 B14 How do you define the realism of the tool design? 4.5 0.8 B15 How do you define the usefulness of interface design for reading measurement? 5.8 0.4 (5) Step #5 : clinome ter tool B16 How do you define the usefulness of tool instruction at screen? 5.6 0.8 B17 How do you define the intuitiveness of the tool usage? 6.2 0.5 B18 How do you define the realism of the tool design? 5.3 0.9 B19 How do you define the usefulness of interface design for reading measurement? 4.9 1.2 (6) Step #6 : reporting note tool B20 How do you define the usefulness of tool instruction at screen? 4.7 1.4 B21 How do you define the intuitiveness of the tool usage? 2.5 1.2 (7) Step #7 : ite ration of experiments over target area B22 How do you define the easiness of finding area centers by location tool? 3.1 0.8 B23 How do you define your learning proficiency as iterating experiments through areas 4.2 0.9
58 Table 2 4 List of quest ions and scores of evaluation form: Question Category C Item Questions Mean Std (1) User overall experience of virtual experiment C1 How do you evaluate the VLF instruction taken before conducting the virtual experiment? 3.5 0.9 C 2 How long did you t ake to complete first measurements at first experiment area? (sec) 523.0 23.5 C 3 How long did you take to complete last measurements at last experiment area? (sec) 396.0 19.8 C 4 How much do you define the effect of computer performance? 4.5 0.8 C 5 How much do you define the easiness of controlling virtual instruments? 3.2 0.9 C 6 How much difference do you feel between your expectations for VLF and actual implementation? 4.8 0.7 C 7 How much do you define the completeness of the VLF system? 4.4 1.1 C 8 How much do you define the realism of the virtual contents? 3.7 1.2 C 9 How much do you define the value of system supports for understanding the VLF? 3.8 0.9 C 10 How much are you satisfied with the VLF? 4.3 1.5 Table 2 5 List of actual and measured tree height and DBH at experiment sector A. Tree Number Tree Height ( m ) DBH ( cm ) Actual Value Measured Value Actual Value Measured Value Mean Stdev Mean Stdev 12 15.24 14.58 2.216 14.48 14.08 1.761 35 16.46 16.01 1.950 22.61 23.10 1.974 36 16.76 17.31 2.488 11.94 12.89 2.104 49 17.07 17.24 2.123 16.51 16.34 2.554 52 17.09 16.47 2.274 19.81 19.16 1.760 53 17.37 16.79 2.433 13.46 13.96 1.911 71 17.68 18.41 2.265 19.30 20.33 2.113 86 18.29 18.89 2.868 23.62 24.34 1.762 94 18.59 18.10 2 .045 23.37 24.43 1.950 96 18.90 19.55 2.368 16.51 17.48 1.635 108 19.24 18.74 2.637 17.27 16.93 1.628 113 19.20 19.72 2.062 22.86 23.14 1.867 118 19.51 19.27 2.271 20.83 21.60 2.203 138 19.81 20.19 2.336 35.05 35.47 2.010 143 20.12 19.64 2.477 29.05 20.06 1.649 147 20.22 20.92 2.023 24.13 23.98 2.368 173 20.63 20.27 2.145 23.37 24.36 2.128 174 20.73 20.90 2.333 24.38 25.39 1.914 355 24.08 23.68 2.033 30.99 31.96 1.838
59 Table 2 6 List of actual and measured tree height and DBH at experiment sect or B. Tree Number Tree Height ( m ) DBH ( cm ) Actual Value Measured Value Actual Value Measured Value Mean Stdev Mean Stdev 209 21.34 21.11 1.850 30.99 30.75 1.708 211 21.64 21.08 1.705 21.34 22.05 1.408 218 21.44 21.88 1.847 27.69 28.04 1. 566 259 22.25 22.16 1.732 32.26 32.68 1.517 277 22.56 23.12 1.894 28.70 28.24 1.442 279 22.56 21.89 1.790 29.72 28.74 1.721 282 22.56 23.52 1.968 30.48 30.12 1.444 298 23.16 22.96 1.809 26.42 26.90 1.058 315 23.47 23.78 1.496 21.59 22.02 1.186 317 2 3.47 23.13 1.821 24.64 23.99 1.827 347 24.08 23.78 1.736 24.89 25.66 1.424 375 24.69 24.90 1.732 30.99 31.25 1.319 383 25.30 24.62 1.639 33.02 33.93 1.548 385 25.60 25.18 1.543 27.18 26.64 1.852 388 25.60 25.84 1.847 33.02 32.96 1.278 393 25.91 25.56 1.413 33.22 33.97 1.558 Table 2 7 Statistical values representing errors and measurement agreements Agreement Tree Height Diameter (DBH) Sector A Sector B Sector A Sector B MSE (1) 0.263 0.216 0.556 0.334 RMSE ( 2 ) 0.513 0.465 0.746 0.465 MAE ( 3 ) 0 .486 0.410 0.677 0.523 EF ( 4 ) 0.905 0.931 0.977 0.982 INDEX ( 5 ) 0.975 0.983 0.994 0.996 (1) Mean Squared Error (2) Root Mean Squared Error (3) Mean Absolute Error (4) Nash Sutcliffee model Efficiency (5) Willmott Agreement Index
60 Figure 2 1 An example of conceptualization of forest learning system. It represents a system concept diagram.
61 Figure 2 2 An example of conceptualization of a forest learning system. It represents system interactions and events.
62 Figure 2 3 Virtual Learning Forest (VLF) world Client/Server architecture showing independent and extendable relationships between VWE and d eveloped virtual world elements
63 Figure 2 4 A virtual object of the sampling area tool. It consists of a grap hic model (center pole, tree name tag, area boundary) and a rendering model to ha ndle boundary size and location
64 Figure 2 5 A virtual object of the logger tape. It contains a graphic model for tape, a process to define distance from target, and a rendering model to handle changes o f shape by moving position
65 Figure 2 6 A virtual object of the clinometer which is used to measure the height of tree. It contains a process to define measurement based on angle of view.
66 Figure 2 7 A vir tual object of the diameter tape. Diameter is calculated by intersecting a grid object to a tree object, and it results in showing diameter value in display and appearing tape line around tree
67 Figure 2 8 The Welcome Board that is displayed at th e start of each session
68 Figure 2 9 The Tool Bag of the instruments that are available for use in the VLF.
69 Figure 2 10 A n user interface object of User Guide
70 Figure 2 11 The Logger Note used by a student to record tree measurements m ade with the instruments.
71 Figure 2 12 An example of enhancement of user interface design and interaction: Diameter tool (a) before modification, and (b) after modification
72 Figure 2 13 Locations of two experiment sectors in the virtual f orest
73 CHAPTER 3 VIRTUAL WORLD DESIGN APPROACH FOR ABSTRACTING AND CHARACTERIZING BASIC VIRTUAL OBJECTS Background In the previous chapter a prototype VLF is developed in the traditional way of implementation by manual coding following the propr ietary model protocol of the VWE. Three types of virtual objects are developed, which are 1) terrain object, 2) forest object and 3) instrument object. A virtual object of terrain is an object focusing on a graphic model and its geological information, whi le a forest object is a combined model of the graphic model of a tree and dynamic model simulating size and spatial distribution of a population of trees by generating geological and physical information for each tree. Virtual objects of instruments are ob jects which stress their behaviors including interactions with the controller (avatar) and communication with target objects as well as a graphic model of instrument shape and an interface for representing data. With the traditional development methods, v irtual objects are expressed in the form of structures oriented in the VWE, which wrap complicate d information of relationships between graphic models, behaviors and processes into a proprietary format. But, it results in an inflexible structure and makes it difficult to create a new virtual world or modify processes in a virtual object. Behaviors of virtual objects are implemented by using programming languages and expressions required by the specific VWE, so it is difficult to understand and share them ef ficiently. Processes (logics) of calculation within virtual objects like instruments can be represented in an independent form like mathematical expressions to enhance extendibility of model by using tag based expressions or web based services.
74 In this ch apter some issues of proprietary model complexities will be described, includ ing model structure, data flow and graphic format. And, it will be shown that flexibility of virtual objects can be enhanced by abstraction of virtual objects implemented, in par i or and model processes can be represented in an abstract, platform independent manor Complexities of Proprietary Model for VWE In the previous chapter, the prototype VLF was developed. The development of virtual o bjects follow s the proprietary format of OpenWonderland. It causes some limitations of model structure, data flow and graphic model to reuse the objects in different virtual worlds and to share them with other VWEs. Model Structure OpenWonderland provides a proprietary way to develop virtual objects. To represent detail, object specific behavior or process it is usually implemented as a module object, a virtual object. It is efficient to develop a complex model by using methods supported in the specific VW E, but it causes difficulties for sharing information and representing behavior and process clearly, or moving the virtual objects to other VWE platforms. That usually requires that the virtual objects be rewritten entirely in the language required by the platform. In the development of the prototype virtual world, environment objects (terrain and forest) are implemented as module objects, and instrument objects are designed to use additional HUD components in addition to a graphic model representing a dev ice. They are represented in solid, concrete formats, in which behavior, process and graphic model are mixed complicatedly It is difficult to distinguish them from fundamental structure explicitly.
75 For example, to show the pattern of plug in structures, two class diagrams of virtual objects are shown in Fig. 3 1 and 3 2 Fig. 3 1 illustrates classes which represent a forest virtual object following the proprietary format of OpenWonderland. Colored classes are required classes for representing a forest mod ule in the VWE. PineForestCellRenderer class includes a description of the tree graphic model resource, forest generation process and dynamic LOD tree model control, which are represented in external classes. Terrain object and graphic models representing instruments use similar format with the forest object. In the OpenWonderland, a virtual object can have a window based interface, a HUD component, which can be used for presenting data. Fig. 3 2 shows the class structure which represents a HUD object for an instrument (diameter tape) in a proprietary OpenWonderland format. A HUD object can include classes for view manager, window decoration, and manipulation of external virtual object. DiameterJPanel class describes window components of HUD object such as size, position, and contents of HUD window. DiameterTapeEntity class represents graphic entities that are associated with the HUD object. The graphic entities include a diameter geometry model and a n instan ce internal diameter tape around the tree trunk. It describe s the model loading method, location and transformation, and dismissal. The diameter tape object is created with segments of line graphic component. Data Flow It is possible that there are different architectural designs for a virtual world depending on the purpose of the system. A client architecture design for data centered representation of an instrument object is shown in Fig. 3 3 It is utilized for the development of the prototype VLF. Actions of user and objects are represented as sol id
76 lines, and dash lines describe information paths. The 3D instrument object (for sharing appearance) and HUD window (for data representation) are separated for efficient manipulation of shape and data representation respectively. A HUD window is the main controller for a user when the instrument is activated, and a graphic model of object is instantiated and attached to an avatar. The measuring process is invoked by user's action such as a picking location on the target object to be measured. Location da ta are sent to the HUD window in the client side, data is calculated, translated, and presented by following a pre defined interface specified by a developer. The user can utilize a rich representation by using the HUD window for his/her own use, but the c lient side requires many resources for data representation. An o ntology is a way to map the abstracted data and relations into a data structure, and it capture s data in complicate d relationship s The ontology can be used to create data structure s for a vi rtual object. A model integrates graphic model, data representation and logic in the instrument virtual object, and manipulates information through a context interface communicating with an ontology manager on the server side. Required resources in the cli ent side can be reduced and an object can be formed by utilizing a highly abstract process. What is needed is work on reducing the burden on the server, enhancing data representation, and developing a generic, rich ontology for interoperability. Fig. 3 4 shows a client architecture design for a model centered representation of instrument objects. It is an alternative to the data centered representation. It focuses on the role of the model rather than data flow. Although the data centered representation
77 sho ws that each object such as a HUD component and virtual object model manages data and relation inside of them, the structure of data control is complex and it is difficult to build the data flow that is independent to the based VWE. Meanwhile, a model cen tered representation gather s related information in the model and simplifying data flow. Graphic Model Format Many VWEs are restricted in support of multiple graphic formats, because they are based on a specific graphic engine. OpenWonderland supports gr aphic formats of X3D and COLLADA, while Croquet utilized a OGRE (Object oriented Graphic Rendering Engine www. ogre 3d.org ) file format and Second Life has its own graphic elements which can be used internally. A graphic format contains data for graphic el ement structure and graphic engine specific attributes for controlling graphic elements and effects such as animation and scripts. And, it is usually formed as a runtime object (a binary file), so it need a specific program library to interpret its content s. It can be converted into other formats by some Digital Contents Creation (DCC) software such as Blender, but there are limitations on file format which can be supported. The efforts to create a compatible graphic format focused on representing only com mon graphic elements like geometry and material, and COLLADA is an example. It excludes graphic engine oriented attributes from the contents, and It describes the elements in a human readable form such as XML representation. Although its simplified content s description provides limited supports to proprietary attributes, the approach can be utilized to represent a graphic model in a diverse data description form such as ontology which is independent of graphic format and graphic engine.
78 Conceptualization of the VLF Educational Objects The prototype VLF is an e Learning system that attempts to improve student field skill with a educational virtual world. The system includes many complicated, diverse concepts in a virtual world. For an example, a small area of a longleaf pine forest ecosystem was built with a terrain object and longleaf pine trees. The forest imposes information of population, distribution and tree size Educational field instruments were created by mimicking the shape (geometry model) and us e in the real world. As an educational learning system, it populated with many actors such as student, teacher and system manager, and there were many complicated relationships between the actors and their activities against environmental objects (e.g. na ture) and equipment objects (e.g. instrument). By expanding the scope of interests on the system to the educational environment, it is possible to define the role of objects in the system and to capture detail ed features of them (Checkland, 1999). As an ex ample of the systematic approach, Soft System Methodology (SSM) (Checkland, 1999) provides a framework for understand ing the complicate d system situation. SSM is a methodology for analyzing complex problems and is useful for developing system models such as the VLF. It uses iterative development cycles including system requirements to develop the design of the system, and the methodology makes it possible to implement a prototype in a short period. And, with a recursive iteration of analysis, design and im plementation, system requirements are realized and implemented rapidly. Usually, SSM consists of several steps The problem in the domain is described with a rich picture, and possible solutions are defined to improve the problem situation. A conceptual mo del is developed, which is based on the
79 activities of the actors in the domain. And, implementation is followed by recursive comparisons with reality and improvement of the model. In order to conceptualize the role and functionality of virtual objects, som e steps of the SSM are adopted to analyze the Virtual Learning Forest. Domain interests surrounding the VLF are described and the situation is conceptualized. The role and activities of virtual objects are specified by creating some modeling diagrams. Doma in Interests Traditional education methods for teaching field skills of measuring and sampling in forest have faced problems of cost, time and space. It is difficult to conduct laboratory exercise in a wide variety of interesting place s Field experiment requires cost for transportation, as if the field of study may be located far away Even though limitations of cost and field may be resolved, weather and safety may be other limiting factor s C hanges of educational policy or situation stimulate the devel opment of the Virtual L earning F orest system in the school domain. The system consists of a server, virtual forest libraries, a model ontolog y for tree growth and forest generation, and GIS data of the terrain and forest. The s chool domain includes teacher student, system manager and researcher. Other schools and international institutes cooperate to enhance the learning program and asset s of the virtual environment. School or system manager respond to the requirement s of education methodologies according to changes in educational situation, policy, and technical improvement s R equirements from the community determine the direction and content of education in school. Teacher provides educational material to the learning system and sets plans for field expe riments. Student use the learning system for learning and developing field skill through the virtual environment. With responses from student and
80 feedback from other users, researcher evaluates the virtual learning system and learning contents. Results of evaluation are used for improving the learning system and enhancing the learning environment. Conceptualization The problem specification comes from various actors such as the system manager, teacher, student and school. Student want to understand teaching material fully, and teacher may want to have a new system which can be managed easily. The school may need to spend less money to build system s for teaching student. A new system can be built for virtual field experiments overcoming limitations of real fi eld trip s on time and cost Investigation of system requirements leads to identifying necessary activities of the system. It identifies flows of main activities to achieve the goal of the system (service V irtual L earning F orest). Serving the VLF requires aggregat ing related activities like managing the forest ontology, generating the forest, and developing virtual instrument. Model Representation T o specify and characterize virtual objects for the virtual learning forest system s ome processes to constru ct model are needed They are represented as U nified M odeling L anguage (UML) diagrams includ ing use case, sequence, and collaboration diagrams Fig. 3 5 illustrates a use case diagram of the virtual learning forest system, which gives information of actors and activities in the system. Key elements (actors and use cases) are described below.
81 System manager creates virtual forest and virtual instruments and manages them in the learning system Teacher uses the virtual learning system to plan contents of virt ual field experiment Student uses the virtual learning system for learning and developing field skills Researcher evaluates the virtual learning system and contributes to develop diverse virtual instrument for enhancing the capability of teaching student Other institute shares the virtual learning system for teaching and cooperate for enriching the virtual learning environment A sequence diagram of the virtual forest learning system is shown in Fig. 3 6 It includes subjects, objects and sequences descr ibed in the Use Case diagram and instrument objects and tree object are represented to describe required information and activities Student and teacher are two subjects in the diagram. Student use instruments in the order of sequences as described in the diagram. The instruments are represented in order of use in a field experiment, and their activities are numbered in according to the order of subject's use. T wo objects, diameter tape and logger tape, communicate with the tree object to get values of me asurement by using the instruments. Teacher is involve d with two sequences to generate tree s in the virtual world and to evaluate measurements by students. A collaboration diagram (Fig. 3 7 ), an interaction diagram, describe s the structural organization of the objects that send and receive messages in the virtual forest learning system. The s equence diagram and collaboration diagram are semantically equivalent, but they each provide a different aspect. Collaboration diagrams focus on the structure and contr ol patterns of objects and their relations with
82 other objects, while sequence diagrams highlight the time sequence of activities which are useful in describing Use Case Scenarios. Students pass the position information of the instruments to locate the i nstruments (sampling area tool and height measure pole) or to get measurements from the instruments (diameter tape, logger tape, and clinometer). A logger note receives entry values from students. In similar way s some instruments (diameter tape, logger ta pe, and clinometer) use position information to get actual location information from the terrain or tree. Teachers generate a forest by passing simulation information to the terrain and tree models and gets reports of measurements for the student. Teacher gives evaluation result to student and can get responses of student about the learning experiences. Abstraction of Objects into Form of Virtual Objects A virtual world consists of various virtual objects that represent things (e. g tree) or concepts (e. g user's action to target object) that exist in real world. Usually creating a virtual object involves processes of specifying their shape (geometry) and function and generalizing properties and relationships. In this section, virtual objects in the VLF are identified and defined and then their features are described, and it includes objects of tree, logger tape, diameter tape clinometer, and sampling area tool. It is explained how each object works and how it relate s to other processes. And, it is described how the virtual object is created based on the real object and utilized in the virtual world. Tree Virtual Object It is assumed that longleaf pine is the only tree species in the virtual world in order to simplify and reduce the complexity of real world e cosystem. Longleaf pine is a
83 pine native to the southeastern United States (Fig. 3 8). It is used for high value wood products, and it is known that ecosystem of longleaf pine is unique and important to wildlife (Peet and Allard, 1993) A longleaf pine tre e typically reaches a height of 30 35 m and a diameter of 0.7m. The bark is thick, reddish brown, and scaly. The leaves are dark green and needle like, and occur in bundles of three. A longleaf pine tree geometry model (Fig.3 9) was created with SpeedTre e (SpeedTree, 2002) a commercial 3D tree library. The tree digital content wa s modified by reallocating shapes and adjusting resolution of textures to refine the quality level of the model. It consists of trunk, branch and leaf elements, but no root. T he trunk geometry is a cone shaped hollow cylinder, with the cross section being a n irregular circular shape. Element s branch and leaves are represented by a rectangular plane and pyramid shaped plane with highly detailed figure based texture, so there is no cross sectional information about them. The prototype tree model represents a n example of a longleaf pine tree with age between 25 30 years old. Various size d and aged longleaf pines are generated from the prototype tree model by applying transformati ons to the size of the prototype model, and the transformation information is randomly developed based on a longleaf pine height diameter distribution table developed from a natural stand of longleaf pines (Leduc and Goelz, 2009). With the 3D tree digital content, three more different types of tree model s are created to apply the discrete LOD to the tree model. The three 2D plane tree models use an image of longleaf pine as a texture They are set to appear at the range of 0m, 30m, 60m and 90m (Fig. 3 10)
84 B ased on the analysis of tree object, Table 3 1 shows the abstraction of tree virtual object in according to categories of view, control, process and data. The tree object appears in 4 different tree models, the views are controlled by LOD system. A mode l is generated by forest generation process. It uses a calculation process for tree population. With input of tree age data, tree location and shape are created. Clinometer Virtual Object A clinometer is an instrument for measuring the angle of a slope. In forestry it is commonly used to measure slope, vertical angles, and tree heights. For measuring tree heights a clinometer has an optical reading lens representing heights in different scales (percent, distance) by converting from the angle measured and kn owing the distance from the point of observation to the tree being measured There are different types of clinometers for measuring only angle, only height or both. For an example, SUUNTO ( http://www.suunto.com ) prov ides various clinometers used in forestry. The SUUNTO PM 5 series presented diverse clinometer models distinguished with types of scales (degree, percent, and distance) and different clinometer body shapes. As an example, t he SUUNTO Model PM 5/66 PC provid es a scale of angle (degree) on a side window, a percent scale (0 +/ 150 % based on distance to target) at the left side in the optical lens, and a height value scale (1:66, height measured at 66ft distance) at the right side in the optical lens. Fig. 3 11 shows major parts of the Suunto PM 5 clinometer. The optical pin hole provides two scales for reading height instantly and a cross hairline directing measuring value. On the side of the body there is side window which consists of a circular angle scale, hairline for angle, and labels of left/right scale in pinhole. On the opposite side window a conversion table is provided for providing angle, distance or percent instantly.
85 To measure the height of a tree ( e g Suunto PM 5), an observer needs to know the distance from the tree. The observer can read angle or percent of height over the particular distance by aligning the optical pinhole to measuring the angle to the top of the tree. The real height can be calculated by multiplying real distance to the measured height percent value. H = tan(alpha) D = percent/100 D where, alpha is a reading view angle, D is a distance from tree, and percent is a reading percentage value on clinometer scale. A clinometer geometry model is developed in 3D, which repre sents a Suunto PM 5 clinometer (Fig. 3 12 ). It includes a set of geometries such as scale wheel, box frame, pin hole frame, scale guide pane and pin hole screen (Fig. 3 13 ). Scale wheel is in a shape of a cylinder, and it has a n angle scale on the flat side surface and a bar scale on the curved surface. Zero index of side angle scale leans toward the center of the earth always under the influence of gravity. B ased on the analysis of clinometer object, Table 3 2 shows the abstraction of clinometer virtual obj ect in according to categories of view, control, process and data. The object can utilize a HUD interface to show a measurement with a pinhole with two scales, Model appearance control is conducted within the Tool Bag, a HUD interface, and A Pin Hole HUD i nterface is utilized to control the measurement window. There are processes for obtaining view angle from the measurement HUD window and for calculating a height with the angle. Logger Tape Virtual Object A logger tape is an instrument for measuring the di stance between objects or length of a object much like a common household measuring tape For forestry field
86 experiment, it is used to measure the distance from tree in feet or meter scale. Generally, the tape is used for measuring short length s or distan ce up to 100 feet. There are different types of logger tape, but they have similar shape s and simple structure. Most of them consist of a metal or plastic box body and metal (nylon or plastic) strip ruler with linear measurement markings. A self retracting button is a part of box body. A typical m easuring ta p e is capable of measuring down to 1/32 inch (0.079375 mm), and double scales rulers are applied such as inch, feet and meter. In Fig. 3 14, a common measuring tape is shown with a description of major p arts of tape. Measuring a distance starts from attaching the end of metal strip ruler to target object and end by reading the value of the ruler when user stands at target location. In the virtual world, it is complex to implement a sagging effect of tape or represent detail s of ruler in the small face. Simply, without considering a tape sagging effect, a distance from a target to user location can be calculated like below equation. where, D is a distance, TPV is a target 3D posi tion, and SPV is a source 3D position. A geometry model for a logger tape was created in 3D model, for a Stanley Powerlock (Fig. 3 15).It is developed as a set of geometries including tape box body, retracting button and metal strip ruler. B ased on the analysis of Logger Tape object, Table 3 3 shows the abstraction of Logger Tape virtual object in according to categories of view, control, process and data. A graphic model represents a diameter tape, and a tape strip to target object is created
87 as a temp orary object within a VWE. Mouse picking action control s the location of the graphic model, and provides the location to distance calculation process. Diameter Tape Virtual Object A diameter tape (D tape) is a form of dendrometer which consists of a cloth or metal tape and tape box. It is mainly used to measure diameter of the tree at breast height (DBH). Standard DBH is measured at a height of 4.5 feet above the ground because it is known that it is convenient to measure diameter at the height. Diameter ca n be read directly from the scale on the tape because the tape is calibrated in units of 3.14 (PI) inches or centimeters. The measured circumference of tree is transformed in to a diameter by dividing the circumference with PI. The obtained diameter is an approximate value because it is assumed that the tree trunk cross sections are perfect circles. In Fig. 3 16, a diameter tape is shown with a description of major parts, and there is a figure to show how to read the measurement with diameter tape. To measu re the diameter of a tree, the tape is wrapped around the tree at 4.5 feet above ground, and the diameter of the tree is a value at the tape where the number '0' aligns with the rest of the tape. Although the cross section at a measured height usually has no exact circular shape, the tape provides a diameter based on an assumption that the tree trunk has a circular shape. The diameter is a calculated value, and as an example it can be represented by below equation. Dia = CF / PI where, Dia is a diameter, CF is a circumference of tree trunk, and PI is 3.14. A diameter tape geometry model is created in 3D, and it is shown in Fig 3 17. It is a simple tape model which consists of tape box body, retracting button and tape ruler.
88 Based on the analysis of Diameter Tape object, Table 3 4 shows the abstraction of Diameter Tape virtual object in according to categories of view, control, process and data. A temporary model, tape strip, is created around the tree truck, and a HUD interface is utilized to show reading mea surement of diameter and the measuring height. The model is handled through input onto the Tool Bag HUD interface. Mouse picking action provides location data for three calculation processes for circumference of tree trunk, diameter based on the circumfere nce, and measuring height from ground. They results in a measured diameter and height. Sampling Area Virtual Objects : Center pole, Boundary tape, and Tree tag A sampling area is an area with a boundary within which tree measurements are taken (usually fo r all trees within the sampling area). The sampling area is usually designed and planed on paper map or with GIS software in the shape of a grid or circles. There is no magic tool for this work, so many simple tools such as paint marker s map s and GPS are used together. The l ocation of sampling area s can be located by GPS, and boundar ies of the area are marked at the tree trunk by using paint. P lastic tree identification tag s are placed on each tree to be measured for use in identification In Fig.3 18, mat erials and tools used for assigning a sampling area are shown. For educational field experiment s circular sampling areas can be determined with a given center location and radius, and the boundary can be marked with paint on the tree. For example, point locations (e.g. (x,y)= (BPX, BPY) ) on the circle boundary can be generated with below equation.
89 ( BPX SACX ) 2 + ( BPY SACY ) 2 = RAD 2 where, BPX, BPY : V alue s of boundary points SACX, SACY : X, Y value of sampling area center points RAD : Radius of sampling area Similarly to attach tree tags in the sampling area it is ne cessary to locate all trees existing inside of the boundary. They can be found with below formula X = (TLX, TLY) if ( TLX SACX ) 2 + ( TLY SACY ) 2 RAD 2 <= 0 where, X : Tr ee location point TLX, TLY : X, Y value of tree location SACX, SACY : X, Y value of sampling area center points RAD : Radius of sampling area The s ampling area tool consists of three tools : center pole, boundary tape and tree identification tag. Area ce nter pole is created in 3D digital content (Fig. 3 19). Boundary tape and tree identification tag are temporary objects, which change their size, shape and contents as the location of area center changes. Based on the analysis of Sampling Area Tool object Table 3 5 shows the abstraction of Sampling Area Tool virtual object in according to categories of view, control, process and data. A graphic model, Center Pole is utilized to represent the area center, and temporary models like boundary tape and tree ID tags are created internally. A HUD interface is used for entry of sampling area radius. Two calculation processes refer to the location picked by mouse action, which are calculations for boundary tape
90 location intersecting with ground surface away at area radius and for tree ID tag location intersecting with tree within sampling area. Summary This chapter explores ways to abstract and characterize virtual object s in the agriculture and natural resource domain. As an example, the developed VLF system in the previous chapter was analyzed T he role s of objects in the system are captured by adopting a systematic approach, and further detail s features and functionality of objects are obtained by creating modeling diagrams such as use case diagram, sequence diagr am, and collaboration diagram. Virtual objects such as longleaf pine tree, logger tape, diameter tape, clinometer, and sampling area tool are analyzed by abstract ing feature s functionality, relationship s and data. Analysis of proprietary model structures gave ways to indentify properties related with proprietary structure of model and to distinguish the independent properties among others. And, two model representations of data centered and model centered are compared to enhance the data flow structure an d virtual object design. A model centered representation can simplify the data flow, and capture complicate d data structure with ontology. The abstracting and characterizing process can help to indentify the role of elements in the model. The identificatio n information can lead to develop ing flexible and extendible model s The approach abstract s complicate d information about relationships between graphic models, behaviors and processes from a proprietary format
91 Table 3 1 Abstraction of Tree virtual object. Category Items Description View Graphic Model 4 different tree models with different detail of appearance Control Model Forest generation process View LOD system, 4 detail levels Process Calculation Tree population Data Input Tree age Internal Model transformation information Output Tree Location Table 3 2 Abstraction of Clinometer virtual object. Category Items Description View Graphic Model 4 geometry parts, Pin Hole, Scale Wheel, Body, Side Scale HUD Interface Pi n Hole with two scales Control Model Activate/inactivate by Tool Bag, a HUD interface View Pin Hole HUD controlled by Ctrl+Up/Down arrow key Process Input View angle at Pin Hole HUD Calculation H = tan (angle) for scale Data Input View angle Inter nal Unit Output Heights for units Table 3 3 Abstraction of Logger Tape virtual object. Category Items Description View Graphic Model Tape body Temporary Model Tape strip to target Control HUD Interface Reading measurement at tape strip Model Act ivate/inactivate by Tool Bag, a HUD interface View Mouse picking action Process Input Mouse picking location at target Calculation Distance between target and instrument Data Input Measurement location Internal Unit Output Distances for units
92 Table 3 4 Abstraction of Diameter Tape virtual object. Category Items Description View Graphic Model Tape body Temporary Model Tape strip around tree trunk HUD Interface Reading measurement at tape strip and height tape Control Model Activate/i nactivate by Tool Bag, a HUD interface View Mouse picking action Process Input Mouse picking location at target Calculation Circumference of tree trunk Diameter with circumference Measurement height from ground Data Input Mouse picking location Internal C ircumference Output Diameter, Measuring height Table 3 5 Abstraction of Sampling Area Tool virtual object. Category Items Description View Graphic Model Center Pole Temporary Model Boundary tape, Tree ID tag HUD Interface Boundary radi us entry Control Model Activate/inactivate by Tool Bag, a HUD interface View Mouse picking action Process Input Mouse picking location at ground Calculation Boundary tape location intersecting with ground away at area radius Tree ID tag location in tersecting with tree within given area radius Data Input Area radius, Center location Internal Boundary tape location, Tree ID locations Output Tree IDs
93 Figure 3 1 Structure of Classes which represent a forest virtual object following a pro prietary method used for OpenWonderland format Colored classes are necessary for representing a module in the VWE. Figure 3 2 Structure of Classes which represent a HUD object for instrument (diameter) following the required proprietary method used for OpenWonderland format
94 Figure 3 3 A client architecture design for data centered representation of instrument objects. Actions of user and objects are represented as solid line s and dash line s describes information path. Figure 3 4 A c lient architecture design for model centered representation of instrument objects. Actions of user and objects are represented as line s and dash e s line describes information path
95 Figure 3 5 A use case diagram of the V irtual L earning F orest
96 Figu re 3 6 A sequence diagram of the virtual learning forest system, which focuses on the interaction and behaviors between actors and virtual objects
97 Figure 3 7 A collaboration diagram of the virtual learning forest system
98 Figure 3 8 Longlea f pine trees (from Google).
99 Figure 3 9 A longleaf pine tree geometry model created by SketchUp (SketchUp, 2000), a 3D modeling program from Google
100 Figure 3 10 Level of detail of the longleaf pine consisting of 4 tree digital resolutions. Figure 3 11 Major parts of the clinometer (Suunto PM 5 model)
101 Figure 3 12 Clinometer (Suunto PM5 model) digital content created by SketchUp (SketchUp, 2000), a 3D modeling program from Google
102 Figure 3 13 Geometries o f 3D clinometer di gital content Figure 3 14 Major parts of the logger tape and detail of the ruler scale
103 Figure 3 15 Logger tape (Stanley Powerlock) geometry model created by SketchUp (SketchUp, 2000), a 3D modeling program from Google
104 Figure 3 16 Majo r parts of diameter tape (JIM GEM Pocket Diameter Tape) and drawing showing how to read a measurement Figure 3 17 Diameter tape digital content created by SketchUp (SketchUp, 2000), a 3D modeling program from Google
105 Figure 3 18 Materials an d tools used for assigning a sampling area
106 Figure 3 19 Sampling area tools (center pole, boundary tape, tree identification tag)
107 CHAPTER 4 AN ONTOLOGY BASED APPROACH TO IMPROV ING MODEL INTEROPERABILITY Background In the previous chapt ers the prototype VLF was analyzed to build virtual objects that could have more flexible structure and VWE independent format s compared with traditional development methods based on proprietary products. Through further conceptualization and abstraction of the virtual system and virtual objects, distinct descriptions of virtual objects including their role and functionality can be developed Beginning with analysis of a proprietary virtual model structure s it distinguished specification data requirements of instruments with in the model frame work required for specific VWEs. And, comparison of two model implementation techniques showed that the model centered implementation of virtual object with ontology could contribute to simplify the data flow between v irtual object and VWE. Ontologies has been utilized to capture basic elements of graphic contents for specific graphic applications or graph model formats (Kalogerrakis et al., 2006; Dartigues et al., 2007; Niknam and Kemke, 2011). However, these ontologi es were restricted to projecting a graphic model for a specific graphic application onto an ontology. The specific graphic application utilizing a specific graphic engine supports many run time attributes to increase the ability of the graphic model, but m ost attributes are dependent to the graphic engine. Thus, the purpose of using ontology was limited to manage graphic models for sharing them for a specific application. The model interoperability is restricted in the specific domain. A virtual instrument of the VLF is a graphic module rather than a static graphic model, which contains static parts (graphic model) and dynamic parts (behavior and
1 08 interaction). To map geometry and behavior of the module into ontology, it is necessary to consider both the grap hic and behavioral/interactive aspects of the virtual objects. Interoperation of model in the VWE domain requires sharing virtual objects with different VWEs. But, there are problems of graphic model format and implementation of behavior and interaction. In this chapter the various objects in the VLF are analyzed from the standpoint of physical, logical and interactive/behavioral characteristics. V irtual objects' geometry ontolog ies are developed that represents geometry information. Model behavior ontolo g ies are constructed to expand the functionality of the graphic model. And, it is described how the interactive virtual objects can be represented and implemented in the VWE to improve model interoperability and system architecture S ystem flexibility and model interoperability are tested through the results of building the same virtual world automatically on two different VWEs OpenSim ulation and OpenWonderland Model Interoperability Generally, t he term of interoperability is defined as the "ability of t wo or more systems or components to exchange information and to use the information that has been exchanged" in the IEEE Standard Computer Dictionary (IEEE Computer Society, 2001). It is an important aspect of application development for improving the effi ciency of the software system. In the VWE domain, model interoperability can be investigated for two aspects of virtual object, a graphic behavior model and a graphic geometry model. The graphic model behavior describes behavior of the geometry model and i nteraction between model and user. The behaviors and interactions may be represented as a simulation or process. It requires descriptions based on the related logic and process, and an ontology to manage them independent ly from the VWE. A
109 graphic geometry model has a specific file format or description of geometry structure. An exchangeable digital content file format which has an open structure contribute s to sharing them. Ontology based Simulation Model Approach The purpose of utilizing ontologies fo r describing a simulation model is to creat e reusable knowledge that can be shared across different projects and platforms But, the design and implementation of ontologies are different depending on the problem domain. Miller (Miller et al., 2004) create d an ontology for discrete event modeling by capturing concepts with a taxonomy of model structural characterization (e.g. State oriented, Event oriented) and model running mechanism. Although Miller utilized an ontology to represent stochastic models, som e researche r s (Jurisica et al., 2004; Cuske et al. 2005) built ontologies containing static aspect s of entities such as simulation data and simulation governing rule. Beck et al. (2008) utilized an ontology to representing simulation logic and processes us ing objects that are editable equation representations. Graphic model ontologies tend to focus on constructing an ontology for a specific data format and sharing models in the same community. For example, Kalogerrakis et al. (2006) created a graphic conte nt ontologies for X3D graphic format to visualize domain knowledge. They generalized X3D graphic elements as graphic content concepts and mapped them into graphic element ontology objects such as a scene and a node. And, Dartigues et al. (2007) created ont ologies for integrating computer aided design (CAD) and computer aided process planning (CAPP) which are commercial software application s It focused on data exchange. A d omain specific ontology is based on the features of the applications, and a shared on tology is used for mapping the domain specific ontology to other CAD family software programs. Similarly,
110 Niknam and Kemke (2011) built a graphic object ontology for a Java 3D graphic scene graph data structure by capturing general graphic concepts such as shape, size and color. J ava 3D has no specific file format, so concepts of graphic contents came from characteristics of the graphic model rather than the file structure. Beck (2009) created a virtual world environment which manipulates graphic models of virtual objects by using an ontology as a content database. On the other hand, Parisi et al. (2007) utilized an ontology to create an animation model for CAD and product management software (PDM). The animation model used a generated script for animation Exchangeable Digital Content File Format Approach I nteroperability of 3D digital content (3D model s ) has been closely related with particular software packages for creating digital content and with platforms for using the digital content. Digital conte nt creation (DCC) software such as 3ds Max and Maya (www.autodesk.com) has been used to create digital assets for game s virtual world s for education social networking, and simulation s DCC software often use s proprietary file format s to code digital ass ets, and it requires various trans coding tools to export and import such files to and from other DCC programs. (Earnshaw and Vince, 2001). Trans coding tools must constantly be rewritten, as the proprietary file formats are changed. Similarly, gaming en gines, which utilize the files in their platforms, require work to support diverse digital formats (Eck, 2006). C ompetition between different DCC software makes it difficult to adapt file format standards Even when the format is opened source it may be very complex to read and to develop an algorithm for using the file data As us e of 3D digital content has expanded the relation between content and runtime (gaming engine) has bec ome more complex to support especially with
111 advanced techniques like phy sical effects and animation. By separating digital content from runtime (gaming engine) physically and logically, it helps to prevent the content from being embedded i n the proprietary runtime engine code and helps to enable generic APIs which can work wit h many kinds of data (Geroimenko and Chen, 2004; Daly and Brutzman, 2007). But, more complex dynamic behavior is required for interactive applications M ore complex control system s need to combine scripted animation, physical simulation and user control. 3D content must be designed for interactivity not only for the interaction with user, but for the interactivity between the different elements of the virtual world An interoperable file format, COLLADA (COLLAborative Design Activity) (Arnaud and Barnes, 2006), became an industry standard at SIGGRAPH '05, which is an annual conference o n computer graphics sponsored by the ACM (Association for Computing and Machinery). It provides an open standard XML schema for exchanging digital assets between various gr aphics software applications. COLLADA documents cover a wide range of features including geometry, material, light, effect, camera, animation, physics and controller. It supports mesh based geometry which is good for handling large, complex dynamic relatio ns between elements rather than other popular formats such as X3D and VRML which utilize specific application oriented scene graph data structures. A scene graph, a collection of spatial representation nodes in a graph (tree) structure, uses many active ru n time attributes for each 3D rendering engine such as 'switch', 'level of detail', and 'script' which COLLADA format does not define. COLLADA merely defines the necessary data for enabling at any applications.
112 Recently, the COLLADA format was adopted by m any DCC software companies and virtual world platforms such as Second Life, Open Simulator and Open Wonderland because of its interoperability. But the COLLADA format does not focus on supporting simulation application s which requires more complex interact ion between elements. Simulation applications need to use intensive controls for user action by handling run time program s The open structure of the COLLDA format enables schema extension, and it is possible to define required information as subset functi on s of the COLLADA format. For example, the immersive education initiative (iEd) created the open 3D/VR file format by extending COLLADA schema for sharing digital assets with diverse virtual world platforms such as Second Life, Open Simulator, and OpenWon derland (Media Grid, 2010). However, the simulation is limited to animation, and it is insufficient to implement simulations requiring complex interactions between user and objects. Ontology based Model and System Ontologies used for abstracting data and simulation processes are useful to define complex data and relations, as well as complex interaction and behaviors, even when behaviors result from simulation And, m any virtual objects in the agriculture and natural resources domain could be developed as interactive 3D model s Information in the ontology is platform independent. By utilizing the ontology and generator, t he geometry, interaction and behavior of the models can be converted automatically in the forms that are required for the specific VWE. An d, it can be more easily shared than the proprietary, platform specific program code In this section, various objects will be analyzed from the point of view of physical, logical and interactive/behavioral characteristics. V irtual object ontolog ies will be developed that represent both object geometry and behavioral (dynamic) information
113 with some ontology authoring tools A g raphic geometry model generator is created to building a graphic model by using ontologies, and model behavior ontologies are util ized to create graphic model modules expand ing the functionality of the graphic model. And, system flexibility and model interoperability are tested through building t he same virtual world on two different VWEs with generated graphic models and model modul es. Ontology Design Ontologies for digital content and module applications are developed for virtual objects including the longleaf pine tree sampling area tool, and instruments such as diameter tape, logger tape, and clinometer. The model ontology descr ibes geometry information of the objects, while the module application ontology is utilized for representing logical and interactive/behavioral information of the object in the virtual world. Tree ontology A n ontology for representing the 3D digital conten t of the longleaf pine tree is shown in Fig 4 1. I t is a geometry model ontology that is utilized for generating 3D tree models. The full 3D longleaf pine model (LPTree Model) is an instance of class 3dModel, and three simplified plane models. Plane model s consists of planes with a texture of front tree projection, and they are distinguished by the number of planes used for the model. A full 3D longleaf tree model (LPTree Model) has relations with instances of 'Asset', 'Scenes' and P arts'. There are three instances as parts that are 'Leaves', 'Branch', and 'Trunk', and they are related with instances of material and geometries such as 'TrunkMesh'.
114 Geometry of the parts are represented with properties of 'pointArray', 'vertices', 'triangle', and 'mesh'. Th e property 'pointArray' is an ontology object containing float arrays, array id, size of array, count of XYZ coordinates. The property 'vertices' contains its id and source float array id. Similarly the property 'triangles' includes information of triangl e count, source id of vertices, and array of vertex order number. The property 'mesh' ha s id, float array ids, vertices ids, and triangles ids. 'Plane 3D' models have similar structures of ontology objects representing their geometry. The 'TreePalne' is a class representing a part of model geometry, and the 3D plane models, 'LPTree1P Model', 'LPTree2P Model', and 'LPTree3P Model', have different number of plane of geometry to constructing 3D tree. For example, the 'LPTree3P Model' has 3 'TreePlane' objects The 3D tree models are a set of tree models representing different levels of detail of the tree based on the distance from user avatar. They are utilized for applying discrete level s of detail (LOD) in the virtual world. A model behavior ontology (Fig. 4 2) of the longleaf pine is an application ontology that is utilized for forming a module to describe relations with other models in the virtual world. It describes 1) how to manipulate ( LPTree MovementOperation ) the LOD of tree according to the distance between tree and user avatar, and 2) how to calculate the value of the 'DistanceFromAvatar' object which represents the distance with values relative to other object in the virtual world. An instance 'LPTree MovementOperation' of 'MovementOperation' class is utilized to broadcast tree location and the changed avatar location. The 'MovementOperation' class has properties of source object name, target object name, and their coordinates. The values of properties are used to calculate a distance between them.
115 An instance of class Equation, 'DistanceFromAvatar', expresses a equation to calculate the distance with propert y value of the 'LPTree MovementOperation'. The equation is represented with operators (Root, Square, Minus) and operands (TreeLocation, AvatarLo cation). Sampling area tool A sampling area tool model ontology is shown in Fig. 4 3. The sampling area tool is a set of three tools, center pole, boundary tape and tree number tag. A center pole is a 3D geometry model which has a stick shape with 2m lengt h and a 5CM x 5CM rectangular intersection. Its asset s scene and parts of geometry are represented in the instance 'CenterPole Model'. A boundary tape and tree id tag are objects constructed from temporary geometry information that are passed through fro m other object's activity such as collision with the boundary of the sampling area. For example, the boundary tape may have a different diameter, and it is placed on the ground that has an irregular surface height due to irregularities in the surface terra in And, tree id tag s must be allocated to different number s of tree s according to the sampling area. An instance 'BoundaryTape' and an instance 'TreeNumberTag' are instantiated from class '2DModel'. The instance 'BoundaryTape' contains instances of 'Line Seg' representing segments of lines composing the lines of the boundary tape. An instance of 'TreeNumberTag' contains instance 'Tag' of 'Box' shape and instance 'TreeNumber' of 'Text'. A sampling area tool behavior ontology is presented in Fig. 4 4. It des cribes how the geometries of the three tools (center pole, boundary tape, and tree number tag) cooperate each other.
116 An instance 'CenterPole Model' is located by information given from instance 'SATool EquipmentOperation' by detecting a user's mouse picki ng point. It is the center location of the sampling area, and it is utilized to seek points on the ground for generating the boundary tape. An instance 'BTGenerate' contains an equation of an instance 'SeekGroundCollisionPoints', and the equation gives poi nts that place the boundary distance calculated with variable 'SamplingAreaRadius' and variable 'CenterLocation'. Similarly an instance 'TagGenerate' utilizes an equation of instance 'SeekTreeArea' for figure tree locations inside of the given sampling ar ea. The equation is formed similar to the equation of the instance 'SeekGroundCollisionPoints', but it facilitates the 'LessThan' operator instead of the 'Equal' operator to figure that the tree locations are located in the region within the boundary poin ts. Diameter tape A diameter tape model geometry ontology is presented in Fig. 4 5. The geometries of the diameter tape during diameter measuring activity are represented with a 3D model of the diameter tape and a 2D model of the enlarged tape part for rea ding measurement on the tape to make it easier for users to read values An instance 'DiameterTape Model' of class '3DModel' represents a 3D model with information about asset, scenes and parts. There are three geometry parts, instances of 'RetractButton ', 'BoxFrame' and 'StripTape'. And, an instance 'DiameterTape HUD' of 2DModel class represents tapes and rulers on tape with 2D geometries such as boxes and lines. Its properties, 'width' and 'height', contain integer values for HUD window size. It has ass ociations of 'Contains' with 2D geometry objects, 'UpTapeStrip', 'DownTapeStrip', 'Crosshair', and 'UDSacle'. Each object has geometry
117 element such as 'Point' object that represents X Y coordinate. For example, 'UpTapeStrip' showing upper diameter tape str ip has two 'Point' objects, 'USUpLeft' and 'USDownRight', that contains X Y coordinates defining left upside corner and right downside corner of the tape strip. An Instance 'UDScale' describ es an upper tape whose scale starts from 0 and lower tape includin g scales for measuring values. The scale of the lower tape changes dynamically during measuring activity. In Fig. 4 6, a diameter tape model behavior ontology is presented. An instance 'DiameterTape Model' of class '3DModel' utilizes 'DiameterTape Movement Operation' which is an instance of class 'MovementOperation' to obtain a location of the 3D diameter tape model that is assumed to be attached to user avatar. The class MovementOperation' contains properties such as a 'resource' (e.g. a String object of object name to check the location) and 'location' (e.g. a String object of float values describing X Y Z coordinate). An instance 'DTape' detects a location on the tree for measuring diameter using information from instance 'DiameterTape EquipmentOperation of class 'EquipmentOperation'. The class 'EquipmentOperation' is utilized for representing data created by using the instrument (e.g. mouse picking position at tree trunk for measuring tree diameter), and providing the data to related objects (e.g. an e quation object 'DiameterOfCircumference'). It contains properties which include 'resource' (e.g. a String object of action resource name, 'mouse'), 'location' (e.g. a String object of float values describing X Y Z coordinate. The data of 'EquipmentOperatio n' is referred by an equation object. For example, an equation object, 'DiameterOfCircumference', utilizes 'Circumference' object to calculate a diameter.
118 I n this process circumference at the target point is obtained by integrating the length of lines t hat consists of points on the trunk intersecting with a n ideal plane generated at the measuring location. An instance 'DiameterTape HUD' utilizes an equation of instance 'UDScaler'. 'DiameterTape HUD' object has an association relation, 'Update', with the 'UDScaler'. The 'UDScaler' object have an instance of equation, 'DiameterOfCircumference'. The equation calculates diameter of the tree cross section with variables of 'Circumference' and 'PI '(3.14). The scales and calculated diameter are represented in the 2D model. Logger tape A logger tape model geometry ontology is shown in Fig. 4 7. There is a 3D model (LoggerTape Model) and a 2D model (LoggerTape HUD). The geometr y of the logger tape during distance measuring activity are represented with a 3D model of the logger tape and a 2D model of the enlarged tape part for reading measurement s on the tape at a tiny area in the virtual world. An instance 'LoggerTape Model' of class 3DModel contains information on asset, scenes and parts. There are three geometry parts, instances of 'RetractButton', 'BoxFrame' and 'StripTape'. And, an instance 'LoggerTape HUD' of '2DModel' class represents tapes and rulers on the tape with 2D geometries such as boxes and lines. An i nstance 'LRScale' is used for describing scale s o n the left and right side of the tape (different scales) and both scale s of the tape change dynamically during the measuring activity. I n Fig. 4 8 a logger tape model behavior ontology is presented. An instance 'LoggerTape Model' of '3DModel' class obta in the location of the 3D logger tape model that is assumed to be attached to user avatar. An instance 'LTape' detects a location
119 picked by a mouse clicking action from information from instance 'LoggerTape EquipmentOperation'. An instance 'LoggerTap HUD' utilizes an equation of instance 'LRScaler', and the equation calculates a distance from target location from the avatar. The scales and calculated distance are represented in the 2D model. Clinometer A taxonom y of clinometers entities (Fig. 4 9 ) is desig ned to represent and store physical and logical information which characterizes a clinometer for realization in the virtual 3D world. It includes information on physical shape ( c lass 3DModel), behavior ( c lass Operation), user interface ( c lass DisplayInter face), and ways to process interaction and behavior ( c lass Process). Class 3DModel has four subclasses representing physical parts of a clinometer, which are BoxBody ' Pinhole ' SideScale and ConversionTable They are sets of geometric groups forming the geometry of the clinometer. It makes it simple to apply changes in appearance of the 3D model such as changing texture, shape type, and background image. Class Operation represents ways of controlling the instrument. Subclass EquipmentProcess desc ribes how to handle the instrument to get measured values by using it, and has properties specifying action invokers ( e.g. mouse wheel up/down for changing clinometer's view angle) and event processers ( e.g. update view angle value). Subclass MovementProc ess explains how to change the position of the instrument in some direction ( e.g. moving clinometer towards or away from the tree), and describes whether it moves by itself or under control of the user avatar while a djusting the instrument.
120 Class Displa yInterface defines the types of interfaces for displaying information such as enlarged the measuring scale bar and displaying tips for help. Subclass HUD is an interface of heads up display, and subclass Popup is a dialog window. Subclasses of HUD c lass, PinholeScale and SideScale describe the scale displays for p in hole in the bar shape and for the side scale angle. Class Process is a subclass representing the mathematical and functional behaviors of the clinometer and user interface. Subclas s Equation describes mathematical expression s of processes. For example, class AngleToHeight is an equation calculating tree height based on the measured sight angle, and is used to show the tree height in the clinometer pinhole. More detailed ontologie s for the clinometer are developed based on the above taxonomy. They describe how the clinometer 3D graphic object is represented and how it can be manipulated by operation information. An ontology for representing a 2D or 3D geometry of the clin ometer is shown in Fig. 4 10 and the model geometry ontology is utilized to generate a 3D or 2D model. The 2D model contains several parts including box e s circle s and lines for representing the clinometer box, optical pin hole, measurement crosshair, and scale. L eft and right scales are a set of scale lines iterated with scale type ( e.g. degree or percent) and the interval s between lines are based on scale origin location and scale size. An instance of class 3DModel represents geometries of the clinometer in 3D, model boundary, and supplement information (e.g. instance Asset) such as creator and unit of geometry. A 3D model includes points, vertices, triangles and mesh geometr ies for representing parts of the clinometer A mesh is consists of sets of triangles
121 c onstructed by vertices which are identified from a point array. An instance Scenes defines the boundary of the model, and represents an aggregation map of geometry groups or components that are defined as separated sets of objects. The application ontolog y (Fig. 4 11) describes 1) how to update (ClinometerEquipmentOperation) display the scale based on changes of user action (Mousewheelaction), and 2) how to calculate (AngleToHeight) a value (TreeHeight (H)) from result of action change (ex: ViewAngle (alph a)) with distance (Distance) from tree A nd, it also explains 3) how to manipulate (ClinometerMovementOperation) the model according to change of location (AvatarLocation). The ontology is utilized to develop a clinometer model module. An instance AngleT oHeight of class Equation is a mathematical expression constructed with left operand, equal ('=') operator and right operand. The equal operator is considered as an assignment operator, not a relation operator. The r ight operand consists of a Multipli er operator, variable Distance and operator Tangent of variable ViewAngle The TreeHeight variable is updated from the equation AngleToHeight A n instance of class ClinometerApp PM5 has an instance of class MovementOperation ', ClinometerM ovementOperation, as a propert ies for defining the behavior of the graphic model according to changes of position (AvatarLocation). The operation manipulates the location of the graphic model in the virtual scene by detecting the location of the avatar wea ring the instrument. Ontology Implementation Every concept involved with an instrument s geometry, behavior and interaction is formally defined by a n object in the ontology. An ontology of an instrument contains concepts such as graphic model, model behavi or, process and other concepts specific
122 to the instrument. A concept contains taxonomic relationships ( e.g. a '3DClinometerModel' is a member of the class Clinometer ), properties ( e.g. a 3D model has a particular shape), and association with other conce pts ( e.g. a clinometer can contain a 2D model, a clinometer contains equipment and/or movement operation for th e instrument) Authoring tools LyraBrowser (Beck, 2008) and ObjectEditor (Beck, 2007) are graphic user interface tool s for authoring ontology obj ects using the engine of the Lyra OMS (ontology management system) (Beck, 2008). Lyra OMS is a server/client environment based on an object database management system for ontology objects The online editors provide effective ways to collaborate with multi ple authors over remote distance s LyraBrowser and ObjectEditor are low level authoring tool s for visualizing and manipulating the ontology as a node and link style graph diagram. LyraBrowser is modified with enhanced navigating and manipulating functiona lities for manipulating ontology elements by displaying the ontology elements in the shape of a network. SimulationEditor and EquationEditor (Beck et al, 2008) are higher level authoring tools using Lyra OMS enviro n ment. SimulationEditor is a tool to crea te a structure diagram of concepts for building a ontology based simulation of a specific physical system (Kwon et al, 2010b) EquationEditor is a tool used in conjunction with SimulationEditor for specifying the dynamic behavior of a physical system throu gh mathematical equations SimulationEditor and EquationEditor are used for authoring equation type processes describing behaviors of model ( e.g. calculation of distance by
123 moving instrument s ) or relat ing part s of a model ( e.g. calculation of tree heigh t with view angle from scale of the clinometer) Building a n ontology The VLF i nstrument ontolog ies which were described in the previous section are implemented with the LyraBrowser. Class VLF is a superclass of instruments, and the instruments h a ve ass ociations with graphic model classes and movement/equipment operation classes. The graphic model concepts represent 2D and 3D graphic model. The operation concepts relate with model behavior such as MovementOperation (it describes the behavior of a graph ic model as effected by user (avatar) movement) and EquipmentOperation (describes concepts related with measurement action s with an instrument). MovementOperation concepts are used for describing the position of the instrument which is usually attached to the avatar in the virtual world. It has a property for reference object name. For example, clinometer and distance tape are assumed to be attached to an avatar when activated, so the reference object indicate s the avatar. While, EquipmentOperation concept associated with an equation object which is used for calculating measurement s of an instrument. Fig. 4 12 show s the relationship between concepts of instruments in a graph and data table. The graph shows that VLF class has instances of instrume nt such as Clinometer ' Diameter Tape and Logger Tape They are listed in the box of all instances, and one of the instances (Clinometer) is highlighted at top of the list box. VLF class has instrument model classes as subclasses. For example, Cli nometer Model class is highlighted at top of the all subclasses box. The arrow between VLF class and the list box describes the relationship between them. The Clinometer Model
124 class has 4 associations with 2 model class ( VLF DC Clinometer 2D Model ' VLF DC Clinometer 3D Model ), movement operation ( VLF Clinometer MO ) and equipment operation ( VLF Clinometer EO ). In the left property window of Clinometer Model class, names of associations and r elated concepts are represented The VLF Digital Co ntent class is created to represent concepts for a graphic model (Fig. 4 13 ). It has two subclasses which are VLF Digital Content 2D Model class and VLF Digital Content 3D Model class. VLF DC Clinometer 3D Model class, a subclass of VLF Digital Con tent 3D Model is a concept representing graphic elements of the clinometer model. It defines a graphic model with elements of asset (model description), scene (scene information of the model) and shape (model geometry information) (Fig. 4 14 ). The shape c ontains parts of geometry with the part name of the model. For example, VLF Clinometer Shape represents its geometry as groups of box frame (VLF Clinometer Part BoxFrame class), conversion table (VLF Clinometer Part ConversionTable), pin hole (VLF Clinom eter Part Pinhole), and scale wheel (VLF Clinometer Part ScaleWheel). In Fig. 4 15 for example, the graph is shown of the relationship between VLF Clinometer Shape class and its subclasses. Each part has associations with geometric elements such as mesh point array, triangle, and vertices. The m esh concept has properties such as ID, IDs of triangle which consists of the mesh geometry Triangle concept contains the used vertices group ID s and order of vertices The order of vertices is a n array of inte ger that represents a vertex id. 'Vertices' concepts contains an id of 'Point array' concept. The 'Point array' concept has an array of float values which is
125 a member of X Y Z 3D coordinate. And, it contains a count of point (total number of coordinate set s) which is used to figure out an id of point from the array. SimulationEditor is utilized to build an ontology of VLF instrument behavior. The ontology contains a high level of concepts for each instrument which has processes in the form of equation s I n Fig. 4 16 four concepts for each instrument (clinometer, logger tape, diameter tape, and sampling tool) are created and shown in SimulationEditor. They are defined as subclasses of SimulationEditor diagram class, and each class is associated with equatio n objects. EquationEditor facilitates build ing and manag ing equation objects within a graphic user interface. An equation object is created by assembling operators and operands. Representing an equation expression starts place the Equal ('=') operator as a root element which allows two operands at left/right sides. Assembled operators and operands construct a hierarchical structure. A left side operand of the Equal operator is restricted to symbols (concepts), while any number or symbols are allowed to be utilized in operators. The equation is decomposed in a hierarch ical order, and each node (operator and symbol) in the decomposition becomes an ontology object with relationships to other operators and/or symbols. In the ontology, EquationEditor element cla ss has two subclasses, EquationEditor symbol and EquationEditor equation that contain symbol and/or equation objects In Fig. 4 17 EquationEditor element class is shown at LyraBrowser. At left window, two subclasses of EquationEditor element class, Equat ionEditor equation and EquationEditor symbol, are listed as a subclass of Equation element class.
126 In the middle window, there are subclasses of EquationEditor equation class which is selected in the left window. The VLF Instrument equation class in the su bclass list is created and utilized for the VLF project, and on the right window it is shown the related information of the class. An equation object is defined as a subclass of the class. The EquationEditor symbol class has a similar subclass relationsh ip with symbol objects. In Fig. 4 18 it show s an equation representing a formula to calculate tree height with given symbols of view angle in percentage (e.g. View Angle Percentage) and distance (e.g. Distance) within the Equation tab of the EquationEdito r. The equation object has a unique equation ID, EQUATION Tree Height from Percentage The equation is created in a hierarchical order. The equal operator ('=') is placed, which has two blank slots at left/right side. The symbol (e.g. Tree Height) repres enting the equation result is set by selecting it among the symbol list, and on the right side blank slot the multiply operator ('X') is assigned. It also has two blank slots at left/right side. On the right side, the symbol, 'Distance' is placed, and at l eft side blank the divide operator ('/') is set. The denominator of the operator is '100', and the numerator is the symbol, 'View Angle Percentage'. The operators and symbols are stored as object s in the ontology. A symbol object represent a concept contai ning a value. It has properties such as Symbol ID, Symbol sign, definition, and unit. Symbols objects are created and managed in the Symbols tab of the EquationEditor (Fig. 4 19) Initial value of the symbol is defined by specifying a source type in the op tion menu For example, selecting the Equation option of source type means that the symbol is automatically associated with an equation which uses the symbol at the left side operand of Equal operator.
127 Digital Content Generator A graphic model geometry ge nerator and a graphic model module generator are developed, as programs written in Java. They retrieve information about graphic model and graphic model module from model geometry ontology and model behavior ontology through Lyra Graphic model geometry g enerator A graphic model geometry generator is an application written in Java that generates a graphic model by using a graphic model geometry ontology. Currently, it generates a graphic model in the COLLADA format that is acceptable in the OpenWonderland and OpenSimulator though generators could be written for other formats The COLLADA graphic format of digital content is widely used for representing geometry, relations between geometric elements ( e.g. joint), and automated action/behavior of geometry ( e .g. wired bone model). Although it has gain ed much interests as a graphic format because of its flexible and expandable format based on xml tags, it has limited ability to describe interaction and complex behaviors with in and between object s The generato r includes a process to interpret the content of the ontology and to create a COLLADA format graphic file by manipulating values of the graphic model with data in the ontology. By modifying a property value of the ontology it is possible to generate vario us model s The COLLADA file format follows the COLLADA specification version 1.4.1 defining 15 library elements to describe graphic content. Elements are animation, animation_clip, camera, controller, geometry, effect, force_field, image, light, material, node, physics_material, physics_model, physics_scene, and visual_scene.
128 A n example of graphic model generation is presented with a clinometer model. Fig. 4 20 shows an example of a geometry model which is stored in the ontology, and it is a scale wheel pa rt of clinometer model. It has a cylindrical shape and contains image textures for ruler. T he geometry model of clinometer was generated with data in the instrument geometry ontology. For example, the Fig. 4 20 showed instances which contain geometry info rmation of clinometer scale wheel. There are 4 instances of mesh, triangle, vertices, and point array. The geometry structure is based on mesh constructed by triangles. A triangle is represented by vertices which is a XYZ coordinates. A point array contain s float values which indicate one value of XYZ coordinates. At F ig. 4 20 (a), instances of 'VLF 3D Mesh' are displayed, which are used for the scale wheel part. The instance is named with the identification number of the mesh after prefix 'VLF M ID'. It ha s properties of 'VLF MSID' ( id number of the mesh), 'VLF P SRC IDS' (id numbers of point array instances referred by the mesh instance), 'VLF V SRC IDS' (id numbers of vertices instances referred by the mesh instance), and 'VLF T SRC IDS' (id numbers of tr iangle instances referred by the mesh instance). For example, a mesh instance, 'VLF M ID8' has an id number (8), and refers instances of point array (id 14, 15, and 17), an instance of vertices (id 16), and instances of triangle (id 1 and 2). It means tha t the mesh geometry consists of two triangles which use a vertices constructed with 3 point arrays. Fig. 4 20 (b) shows a list of triangle instances. The instance is named with the id number of the triangle after prefix 'VLF T ID'. It has properties of 'VL F TID' ( id number of the triangle), 'VLF TCOUNT' (count of triangles), 'VLF MID' (id numbers of material),
129 'VLF VERTEXID' (id number of vertices instance), 'VLF TEXCOORDID' (id numbers of texture coordinate), and 'VLF TPIDS (coordinate order numbers for va lues in a point array instance). For example, a triangle instance, 'VLF T ID1' has an id number (1), which is built with 48 triangles and a material (Material 13). The triangles are represented with coordinates of vertices (ID16), and the order of vertices are shown in order number array (.., 3 3 1 1 3 3 4 4 4 4 ..). A list of vertices instances are shown at Fig. 4 20 (c). The instance is named with the id number of the vertices after prefix 'VLF V ID'. It has properties of 'VLF VID' ( id of the vertices), VLF PSRCID' (id of point array instance representing topology), and 'VLF NSRCID' (id of point array instance representing normal vector). For example, a vertices instance, 'VLF V ID16' has an id (ID 16), and its topology is represented with a point array i nstance (ID 14). The normal vector is created with a point array instance (ID 15). At last, a list of point array instances are shown at Fig. 4 20 (d). The instance is named with the id number of the point array after prefix 'VLF P ID'. It has properties o f 'VLF SID' ( id of the point array source), 'VLF FAID' (id of float values array containing values of coordinates), 'VLF FLOATCOUNT' (count of float values in the float array), 'VLF FLOATARRAY' (array of float values), 'VLF STRIDE' (count of coordinate ele ments, e.g. 3 for XYZ system), and 'VLF SCOUNT' (count of coordinates which are parsed with the given stride). For example, a point array instance, 'VLF P ID14' has an id (ID14), It contains a float array source (ID23) which has 300 float values (.., 0.3773 541459283267 0 0.0842365764077099 0.4169600892469052 ,..). They are parsed as a coordinate of XYZ system (stride 3), and there are 100 coordinates.
130 During the generation, a empty COLLADA format file is created. The, each mesh object of the part in the ontol ogy is collected, and transformed as a 'mesh' node within a 'geometry' node under 'library_geometries' node. Under the' mesh' nodes, data in 'pointArray' object are formed as a source node with id, length of array, and float array. And, data in 'vertices' ontology object are generated into the 'vertices' node under 'mesh' node with its id and source id. Similarly, data in 'triangles' object are formed in the 'triangles' node with order array of vertices consisting the triangles. The result COLLADA file is shown in the Fig. 4 21. Graphic model module generator The graphic model module, a model package set of digital graphics and program code, is used to enrich functionality of static graphic elements The program code is required to be written in a specific programming language dictated by the target VWE. A graphic model module generator was developed in Java and is specific to the particular target platform. Thus there is a generator for OpenWonderland and a slightly different one for OpenSim ulator The generator creates code from the model application ontology, and the code contain s processes that control behaviors of the geometry in responds to user action And, it includes implicit processes of the model for external features which are not included in the digital content ( e.g. a HUD component for measurement indicator of the clinometer). The code is automatically generated by following the API of a graphic library which is embedded in the virtual environment system. For example, a model module written in Java is utilized for the OpenWonderland VWE, while the OpenSimulator VWE uses modules written in C++ and Linden Script Language (LSL) for Second Life.
131 A graphic model module is generated for the OpenWonderland VWE. It is a zip file containing a clinom eter graphic model generated w ith the clinometer model geometry ontology and program codes created with the model behavior ontology The clinometer graphic model is generated in COLLADA. The generator retrieve s model information for the user interface (Cl ass DisplayInterface), geometry (Class 3DModel), behavior (Class Operation) and process es (Class Process) without a ny restriction from the graph ic format of model itself And, it generates program code matching with the VWE platforms for the user interface (a HUD action) and processes (calculating view angle and values of scale ruler) automatically (Fig. 4 22 ). T he model generator creat e s digital content for the 3D clinometer mod el in a COLLADA format. By utilizing the ontology and generator, t he geometry, interaction and behavior of the models can be converted automatically in the forms that are required for the specific VWE. The information in the ontology is platform independ ent, and it can be more easily shared than the proprietary, platform specific program code. Model Interoperability Many VWEs are developed with proprietary programming language, and different graphic engines are used for different languages to provide th e graphic rendering of the system. The proprietary nature of these languages, as well as lack of transparency at the code level limits the usability of digital content and programming code implementing behavior and process and limits ability to move mode ls between different VWEs. For example, the OpenWonderland uses the jMonkeyEngine ( http://jmonkeyengine.com ) and provides a COLLADA graphic file loader to load the file
132 format instantly. And, a model module needs t o contain program code written in a specific programming language Java While, the Second Life ( http://secondlife.com ) developed a graphic engine with OpenGL ( http://www.opengl.o rg ) API, and its specific graphic model is provided to build content in the world. Recently, a COLLADA graphic format model is also allowed to be used in the Second Life. Linden Script language (LSL) is used to control the behavior of objects. COLLADA gr aphic format is supported for generating a graphic model, and Java code and LSL code are provided for the OpenWonderland and Second Life (similar VWE such as the OpenSimulator). Model interoperability was tested by generating and applying a clinometer mode l and forest which can be used in two different VWE s, the OpenWonderland and the OpenSimulator T he OpenSimulator ( http://opensimulator.org ) is a virtual world environment based on the reverse engineered the Second Life s erver ( http://secondlife.com ) As an open source multi platform, multi user 3D application server, it received much attention because of its compatibility with the client for Second Life, and it has a large use r community. A m odel geometry generator created a tree and clinometer 3D graphic model in the COLLADA format. Module generator s are utilized for generating program code for a forest, tree and clinometer modules for implement ing behavior and processes of t he model s. For the OpenWonderland, the module contains processes for loading a graphic model into the VWE repository, scaling and positioning of the model, and linking between parts of the model. The module is generated in the Java language, compiled and packaged in a jar file F or the OpenSimulator, LSL scripts are generated, and they
133 are associated with the graphic model manually and utilized in the system without any compiling procedure (scripts are interpreted) In the module, a function is reserved for detecting interaction with the user interface. And, there is a function to implement a dynamic scale indicator representing measured values converted by calculation functions which are generated from the equation s in the behavior ontology The equation generation is facilitated by a generator application utilizing an equation ontology and object in Lyra (Beck, 2009). In Fig. 4 23 and 4 24 it is shown that a forest of longleaf pines and a clinometer model are placed and activated on two different virtua l world servers, the OpenSimulator VWE and OpenWonderland VWE. T he world s are viewed through the Second Life client program and OpenWonderland embedded viewer, respectively The generated 3D models of longleaf pine and clinometer are uploaded to the VWE, and a forest module application is utilized to populate tree objects in the world. A forest of 100 longleaf pines is created on an small island at the OpenSimulator server (Fig.4 23 (a)), and same number of trees are generated in the OpenWonderland server (Fig. 4.23 (b)). T he generated tree modules for both VWE populate a forest of longleaf pines with given population model information. But, terrains used in the test are directly adopted with a simple model created with the VWE's digital content creation t ool (for the OpenSimulator) and a simple plain model (for the OpenWonderland). T he generated clinometer model is deployed in both VWE s is shown in Fig. 4.24 (a) (at the OpenSimulator) and Fig. 4.24 (b) (at the OpenWonderland). Behavior (changing view angl e) and processes (calculate tree height from angle) of the
134 clinometers are generated in the LSL code and Java code for the OpenSimulator and the OpenWonderland VWE respectively As a result, it was possible to interoperate 3D models with their behavior and the virtual world in other VWEs and avoid problems with the proprietary languages used on each specific VWE. Summary This chapter explores ways to improve model interoperability by using an ontology based approach. The model interoperability in the VWE domain means that a graphic model or graphic module containing processes such as behaviors and interactions can be reused at different VWEs. Ontology may contribute to manage domain knowledge of model behavior and interaction, and geometry of model within forms independent to graphic format and VWEs. Based on results, in the previous chapter, of virtual objects in the VLF created with traditional proprietary method, ontologies for various virtual objects are developed with the point of view of physical, lo gical and interactive/behavioral characteristics. The virtual objects includes tree, logger tape, diameter tape, sampling area tool, clinometer. To describe the characteristics of a virtual model, a model geometry ontology and a model behavior ontology are created for each objects. G eometry data in the ontology is represented with elements of mesh based geometry such vertices, triangle and mesh. A behavior ontology contains concepts of interaction between user and object, and processes (simulation) as a col lection of equation objects. Two generators are developed for creating a graphic geometry file and a graphic module containing graphic model and program code for processes. The graphic geometry generator retrieves geometry data from the ontology, and gener ates a graphic file for the target VWE. COLLADA file formation is supported currently. The module
135 generator creates a module of graphic model and program code, which expand the ability to handling the graphic model and controlling related processes implici tly. System flexibility and model interoperability are tested by building the same virtual world in two different VWEs with generated graphic models and model modules. OpenWonderland and OpenSimulator are the testing VWEs, and the worlds contains virtual tree and virtual clinometer. Graphic geometry models of tree and clinometer are generated in COLLADA file. Modules of tree and clinometer are generated for OpenWonderland, while for OpenSimulator LSL codes are created. It was possible to interoperate 3D mo dels with their behavior and the virtual world in other VWEs, and avoid problems with the proprietary languages used on each specific VWE. Information in the ontology is platform independent. By utilizing the ontology and generator, t he geometry, interacti on and behavior of the models can be converted automatically in the forms that are required for the specific VWE. And, it can be more easily shared than the proprietary, platform specific program code
136 Figure 4 1 A diagram of the longleaf pine model geometry ontology. Gray solid lined rectangles are classes, and orange dash lined rectangles are instances. Blue circles are properties. Arrows show relations
137 Figure 4 2 A diagram of the longleaf pine model behavior ontology. Gray solid lined rectangles are classes, and orange dash lined rectangles are instances. Blue circles are properties. Arrows show relations
138 Figure 4 3 A diagram of sampling area tool model geometry ontology. Gray solid lined rectangles are classes, and orange dash lin ed rectangles are instances. Blue circles are properties. Arrows show relations
139 Figure 4 4 A diagram of sampling area tool model behavior ontology. Gray solid lined rectangles represent classes, and orange dash lined rectangles means instances. Blue circles represent properties. Arrows show relations
140 Figure 4 5 A diagram of diameter tape model geometry ontology. Gray solid lined rectangles are classes, and orange dash lined rectangles are instances. Blue circles are properties. Arrows show rel ations
141 Figure 4 6 A diagram of diameter tape model behavior ontology. Gray solid lined rectangles are classes, and orange dash lined rectangles are instances. Blue circles are properties. Arrows show relations
142 Figure 4 7 A diagram of logger t ape model geometry ontology. Gray solid lined rectangles are classes, and orange dash lined rectangles are instances. Blue circles are properties. Arrows show relations
143 Figure 4 8 A diagram of logger tape model behavior ontology. Gray solid lined re ctangles represent classes, and orange dash lined rectangles means instances. Blue circles represent properties. Arrows show relations
144 Figure 4 9 Taxonomy of the virtual object (clinometer) domain ontology
145 Figure 4 10 A diagram of clinomet er model geometry ontology. Gray solid lined rectangles are classes, and orange dash lined rectangles are instances. Blue circles are properties. Arrows show relations
146 Figure 4 11 A diagram of clinometer model behavior ontology. Gray solid lined rec tangles represent classes, and orange dash lined rectangles means instances. Blue circles represent properties. Arrows show relations
147 Figure 4 12 A n i mplementation example of an instrument (clinomet er) ontology within LyraBrowser
148 Figure 4 13 V LF Digital Content class and its subclasses Figure 4 14 Taxonomical hierarchy of VLF DC Clinometer 3D Model class
149 Figure 4 15 Subclasses and associations of VLF Clinometer Shape class
150 Figure 4 16 Example of building a group of model b ehaviors for VLF instruments with the SimulationEditor an ontology based simulation authoring tool Figure 4 17 Representation of Equation element class and their subclasses in the LyraBrowser. VLF Instrument Equation class is defined as a subclass of EquationEditor equation class which is subclass of EquationEditor element
151 Figure 4 18 A n equation object for representing the tree height calculation with view angle in percentage and distance in equation tab of EquationEditor Figure 4 19 An example of defining a symbol (SYMBOL Tree Height) within Symbols tab in the EquationEditor
152 Figure 4 20 A n example of geometry instances in ontology representing a scale wheel part of an instrument (clinomet er) within LyraBrowser
153 Figure 4 21 An example of generated COLLADA code for a scale wheel part of clinometer model
154 Figure 4 22 An example diagram of Clinometer model module generated from the ontolog ies
155 (a) (b) Figure 4 23 A simple example of 3D model interoperabi lity. Longleaf pine forest model is loaded in virtual world server s: (a) Open Simulator a nd (b) OpenWonderland
156 (a) (b) Figure 4 24 A simple example of 3D model interoperability. Clinometer model is loaded in virtual world server s: (a) Open Simulat or a nd (b) OpenWonderland
157 CHAPTER 5 CONCLUSIONS, CONTRIBUTIONS, AND FUTU R E DIRECTIONS Conclusions Presenting a Virtual World in the Agriculture and Natural Resource Domain The presented work shows that the virtual world environment can be used for educati onal experiment s and simulation. A virtual learning forest (VLF) was developed by using virtual world technology to enhance education experiences without restrictions on time, distance cost or safety VLF consists of three scene components (tree, terrain, and forest) and virtual instruments for mensuration including a diameter tape, logger tape and clinometer. E valuations were conducted including an expert test, performance test, usability test for system design and learning efficiency. The Usability evalu ation gave reasonable acceptance and satisfaction for the system, and the result of learning efficiency test showed that students improved their skills by reducing measurement errors. For forestry education, the VLF shows much potential for offering immers ive learning experience that improve on traditional computer based training. It can offer a laboratory experience that complement s real laboratories which may be expensive, dangerous, or inaccessible to students. The developed of a new forestry educational delivery paradigm, such as the VLF, can overcome limitations on cost, time and location Abstracting and Characterizing Virtual Object s The prototype VLF system was analyzed T he role s of objects in the system are captured, and further detail ed features and functionality of objects are obtained by creating modeling diagrams such as use case diagram, sequence diagram, and collaboration diagram. Virtual objects such as longleaf pine tree, logger tape, diameter
158 tape, clinometer, and sampling area tool are a nalyzed by comparing with real object to abstract feature, functionality, relationship s and data. Analysis of proprietary model structures gave ways to indentify properties related with proprietary model structure and to distinguish the independent proper ties among others. T wo model representations data centered and model centered are compared to represent the data flow structure and virtual object design. A model centered representation can simplify the data flow, and capture complicate d data structure with ontology. The abstracting and characterizing process helped to indentify the role of elements in the model. The identification information lead s to develop ing flexible and extendible model s that incorporate complicate d relationships between graphic m odels, behaviors and processes. Methodology for Enhancing Model Interoperability by U sing An Ontology The presented work shows that an ontology based approach improves model interoperability. The model interoperability in the VWE domain means that a graph ic model or graphic module containing processes such as behaviors and interactions can be reused at different VWEs. Ontology contribute to manage domain knowledge of model behavior and interaction, and geometry of model within forms independent of graphic format and VWEs. Ontologies for various virtual objects are developed with the point of view of physical, logical and interactive/behavioral characteristics. To describe the characteristics of virtual model, a model geometry ontology and a model behavior o ntology are created for each objects. G eometry data in the ontology is represented with elements of mesh based geometry such vertices, triangle and mesh. A behavior ontology contains concepts of interaction between user and object, and processes
159 (simulatio n) as a collection of equation objects. Two generators are developed for creating graphic geometry file and a graphic module containing graphic model and program code for processes. And, system flexibility and model interoperability are tested by building the virtual world on two different VWEs with generated graphic models and model modules. It was possible to interoperate 3D models with their behavior and the virtual world in other VWEs, and avoid problems with the proprietary languages used on each spe cific VWE. Information in the ontology is platform independent. By utilizing the ontology and generator, t he geometry, interaction and behavior of the models can be converted automatically in the forms that are required for the specific VWE. And, it can be more easily shared than the proprietary, platform specific program code Contributions A virtual learning forest system for forestry mensuration exercise s was built It can offer experience that complement real laboratories which may be expensive, dangero us, or inaccessible to students. The new forestry educational delivery paradigm, the VLF, can overcome limitations on cost, time and region Virtual objects in the domain w ere analyzed to abstract and categorize model objects Environment objects and instru ment object can be reused in other virtual world An ontology for virtual instruments was developed It is independent of VWE, and it can be used in automatically generating portable and reusable virtual instruments. Future Directions Expansion of Model In teroperability This study tested two VWEs, OpenWonderland and OpenSimulator, which are two popular open VWEs. Model generation supports only limited graphic format and
160 program code. There are many other VWEs such as Croquet that are utilized for student ed ucation with different graphic format and program code. The ontology based approach to model interoperability can be applied to other VWEs by adopting their requirement into the ontology. Evaluation for Contribution of VLF to Real Education The presented evaluations focused on the measuring learning efficiency with the VLF. It verified that students could improve their virtual field skill with the system by recursive measurement activities. But, it could not cover the effect of the virtual learning into th e real experiments. In future, another evaluation for comparing measurement results in real forest of two groups (a group of students who experienced the VLF versus a group of student who have no experience) may explain the usability of the virtual forest eLearning system for real world education. Digital Forest Learning Library This study focused on improving a virtual world of longleaf pine ecosystem. It consisted of a kind of tree in the world, but it will be necessary to build more realistic ecosystem with diverse plants to give practical learning experience (Barton, 2008) The ontology based methodology can be used to generate diverse forest virtual world by providing characteristic forest population and distribution information to the system. The gene rated forest world can be utilized by other virtual student as a library of virtual forest. Ontology Reasoning The ontology based approach did not focus on ontology reasoning that is valuable function of ontologies. It could be used to selecting appropria te forest model and instrument for different field experiments.
161 APPENDIX EVALUATION FORM OF VLF SYSTEM A. These questions are prepared to understand users and user system. A.1. User's computer resource A.1.1. What is the operating system of your com puter ? A.1.2. How old is your computer? A.2. User's computer ability A.2.1. How long have you played with 3D game? A.2.2. How well do you know about Virtual World? A.2.3. How much have you experienced Virtual Learning Environment? A.3. User's background A.3.1. Did you take a class course related with field experiment? A.4. User's expectation
162 A.4.1. How much do you expect to get positive results on lear n ing after you are instructed about VLF? A.4.2. How much reali ty do you expect to experience before you start VLF? B. These questions are prepared to answer by conducting virtual experiments step by step. STEP #1. User guide & VFL tool box B.1. How do you define the helpfulness of VLF user guide? B.2 How do you define the accessibility of VLF Tool Box? B.3. How do you define the intuitiveness of VLF Tool Box? STEP #2. Measuring area tool B.4. How do you define the usefulness of tool instruction at screen? B.5. How do you define the c orrectness of displaying area center and tree number tags? B.6. How do you define the clearness of area boundary? B.7. How do you define the intuitiveness of the tool usage?
163 B.8. How do you define the realism of the tool? STEP #3. DBH to ol B.9. How do you define the intuitiveness of the tool usage? B.10. How do you define the realism of the tool design? B.11. How do you define the usefulness of interface design for reading measurement? STEP #4. Distance measuring tape tool B.12. How do you define the usefulness of tool instruction at screen? B.13. How do you define the intuitiveness of the tool usage? B.14. How do you define the realism of the tool design? B.15. How do you define the usefulness of interface de sign for reading measurement? STEP #5. Clinometer tool B.16. How do you define the usefulness of tool instruction at screen? B.17. How do you define the intuitiveness of the tool usage?
164 B.18. How do you define the realism of the tool design ? B.19. How do you define the usefulness of interface design for reading measurement? STEP #6. Reporting note tool B.20. How do you define the usefulness of tool instruction at screen? B.21. How do you define the intuitiveness of the tool u sage? STEP #7. Iteration of experiments over target area B.22. How do you define the easiness of finding area centers by location tool? B.23. How do you define your learning proficiency as iterating experiments through areas? C. The se questions are prepared to get answers from user's experience after completing the virtual experiment test. C.1. How do you evaluate the VLF instruction taken before conducting the virtual experiment?
165 C.2. How long did you take to complete first me asurements at first experiment area? C.3. How long did you take to complete last measurements at last experiment area? C.4. How much do you define the effect of computer performance? C.5. How much do you define the easiness of controlling virtu al instruments? C.6. How much difference do you feel between your expectations for VLF and actual implementation? C.7. How much do you define the completeness of the VLF system? C.8. How much do you define the realism of the virtual contents? C.9. How much do you define the value of system supports for understanding the VLF? C.10. How much are you satisf ied with the VLF?
166 LIST OF REFERENCES Ahmed, H. and D. Gracanin.2010. Context Sensitive Interaction Interoperability for Seriou s Virtual Worlds. 2010 Second International conference on Games and Virtual Worlds for Serious Applications, 25 26 March, Braga, Portugal. Anderson, P. 2007. What is Web 2.0? Ideas, technologies and implications for education. Technical report, JISC. Anth opoulos, L., V. Gerogiannis, P. Fitsilis, and A. Kameas. 2010. Training in Profiling, Negotiation and Crisis Management Using and Immersive and Adaptive Environment. CSEDU, 2nd International Conference on Computer Supported Education, 7 10 April, Valencia, Spain. Ann Myers Medical Center. 2010. http://ammc.wordpress.com Arnaud, R. and M.C. Barnes, 2006, COLLADA: Sailing the Gulf of 3D Digital Content Creation, CRC Press, 250 pages, A K Peters, Ltd, Wellesley, MA. Bannister, M.E., 2009, Virtual Learning Forest project, http://www.reeis.usda.gov/web/crisprojectpages/0218300 virtual learning forests developing collaborative opportunities in virtual world environments for undergraduate natural resource students.html ion Weather in Virtual Environments. Internaltional Journal of Computer Game Research. 8(1). Beck, H. W., 2003, Integrating ontologies, object database, and XML for educational content management, Proc. of E Learn., Arizona, USA. Beck, H. W., 2007, Lyra on tology management system, http://orb.at.ufl.edu/ObjectEditor/index.html Beck, H.W., 2008 Evolution of Database Designs for Knowledge Management in Argriculture and Natural Resources, Journal of Information Technology in Agriculture. 3(1). Beck, H.W. 2009. Lyra Virtual World Environment (Lyra VWE): Educational Applications in Agriculture and Natural Resources, 7th World Congress on Computers in Agriculture and Natural Resources Conference Proceed ings, 22 24 June 2009, Reno, Nevada USA. Beck, H.W., K. Morgan, Y. Jung, S. Grunwald and H. Y. Kwon, 2008, Ontology based simulation applied to soil water and nutrient management. In: P. Papajorgji, Editor, Advances in Modeling Agricultural Systems, Spring er, Berlin, pp. 209 242.
167 Bemis Schurtz, H.R. and S. Bussmann. 2009. Fostering New Learning Communities, Nurturing Online Learning Ecosystems. International Symposium on Emerging Technology Applications for Online Learning. June 19. San Francisco, CA. Bisho p, J. 2009. Enhancing the understanding of genres of web based communities: The role of the ecological cognition framework. International Journal of Web Based Communities, 5(1), 4 17. Blender. 1995. http://www.blender .org Boulos, M.N.K. and S. Wheeler. 2007. The emerging Web 2.0 social software: an enabling suite of sociable technologies in health and healthcare education. Health Information and Libraries Journal, 24(1):2 23. Bray, D.A. and Konsynski, B.R., 2007, Vir tual worlds: multi disciplinary research opportunities, ACM SIGMIS Database, 38(4), 17 25. Callaghan, V., M. Gardner, B. Horan, J. Scott, and L. Shen. 2008. A Mixed Reality Teaching and Learning Environment. ICHL, LNCS 5169, pp. 54 65. Campbell, M. 2009. U sing 3D virtual worlds to teach decision making. Same places, different spaces. Proceedings ascilite Auckland 2009, 104 109. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. Checkland, Peter. 1999. Systems Thinking, Systems Pract ice. John Wiley and Sons. COLLADA. 2004. http://collada.org Croquet. 2007. http://opencroquet.org Cranitch, G. and M. Rees. 2009. Virtualisation: A case study in database admini stration laboratory work. Same places, different spaces. Proceedings ascilite Auckland 2009, 168 174. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. Cuske, C., T. Dickopp and S. Seedorf. 2005. JOntoRisk: An Ontology based Pla tform for Knowledge based Simulation Modeling in Financial Risk Management. European Simulation and Modeling Conference 2005. Daly, L. and D. Brutzman. 2007. X3D: Extensible 3D Graphics Standard. Signal Processing Magazine, IEEE 24(6), pp. 130 135. Dartig ues, C., P. Ghodous, M. Gruninger, D. Pallez, and R. Sriram. 2007. CAD/CAPP Integration using Feature Ontology. Concurrent Engineering, 15(2), pp. 237 249. Earnshaw and Vince, J., Editors, 2001. Digital content creation, Springer Verlag, London.
168 Eck, R.V., 2006, Digital Game Based Learning: It's Not Just the Digital Natives Who Are Restless, EDUCAUSE Review, 41(2). Farmville. 2009. Zynga. http://www.farmville.com Freitas, S.D., G. Rebolledo Mendez, F. Liarokapis, G Magoulas, and A. Poulovassilis. 2009. Developing an evaluation methodology for immersive learning experiences in a virtual world. Proc. Of the 2009 Conference in Games and Virtual Worlds for Serious Applications. 43 50. March 23 24, Coventry, UK. Geroime nko, V. and C. Chen, 2004, Visualizing Information Using SVG and X3D, Springer Verlag New York, Inc., Secaucus, NJ. G l. L.F., N. Gu, and A. Williams. 2008. Virtual worlds as a constructivist learning platform: Evaluations of 3D virtual worlds on design te aching and learning. ITcon, Virtual and Augmented Reality in Design and Construction, 13:578 593. http://www.itcon.org/2008/36 Gregory, S. and B Tynan. 2009. Introducing Jass Easterman: My Second Life learning space. Same places, different spaces. Proceedings ascilite Auckland 2009, 377 384. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. Hodge, E.M. and Collins S. 2010. Collaborative Efforts: Teaching and Learning in Virtual Worlds. EDUCAUSE Review. 45(3):62 63. Honey, M.L.L., S. Diener. K. Connor, M. Veltman, and D. Bodily. 2009. Teaching in virtual space: Second Life simulation for haemorrhage management. Same places, different spaces. Proceedings ascilite Auckland 2009, 1222 1224. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. Hudson, K. 2007. Educational Opportunities in Virtual Worlds. Loyalist College, Belleville, Canada. http: //www.slideshare.net/kenny.hubble/edu virtualworlds IEEE Computer Society. 2001. Standard Computer Dictionary: A Compilation of IEEE Standard Computer Glossaries. IEEE Press, Los Alamitos Circle, California. Jack S.B., J.B. Atkinson, M.A. Melvin and J.L Gagnon 2005 C onversion of Immature Slash Pine Plantations to Multi Aged Longleaf Pine Forests over Time: A Conservation Silviculture Approach Proceedings of the Fifth Longleaf Alliance Regional Conference, Longleaf Alliance Report No. 8, May 2005, Hat tiesburg, Mississippi Jarvis, S. and S.D. Freitas. 2009. Evaluation of an Immersive Learning Programme to Support Triage Training. Proc. Of the 2009 Conference in Games and Virtual Worlds for Serious Applications. 117 122. March 23 24, Coventry, UK.
169 Jesch ke, S., A. Gramatke, O. Pfeiffer, C. Thomsen, and T. Richter. 2009. Networked Virtual and Remote Laboratories for Research Collaboration in Natural Sciences and Engineering. 2nd International conference on Adaptive Science & Technology, 14 16 December, GIM PA Campus, Accra. Jung, Y., H.W. Beck, M. Bannister, W.P. Cropper, Jr., and C. Staudhammer, 2010, Virtual Learning Environment: An Application to the Longleaf Pine Ecosystem. ASABE Annual International Meeting. June 20 23. PittsBurgh, PA. Jurisica, I., J. Mylopoulos and E. Yu. 2004. Ontologies for Knowledge Management: An Information Systems Perspective. Knowledge and Information Systems Vol. 6, pp. 380 401. Kalogerrakis,E., S. Christodoulakis, and N. Moumoutzis. 2006. Coupling Ontologies with Graphics Co ntent for Knowledge Driven Visualization. Proceedings of the IEEE Virtual Reality Conference 2006 (IEEE VR '06), PP. 43 50, Virginia, USA, 25 28 March 2006. 22. Kleinschmidt, A. 20 09. Agriculture in Ohio OSU Extension Ag News. Ohio State University Extension. http://agvanwert.wordpress.com Kwon, H. Y., S. Grunwald, H.W. Beck, Y. Jung, S.H. Daroub, T.A. Lang and K.T. Morgan. 2010a. O ntology based simulation of water flow in organic soils applied to Florida sugarcane. Agricultural Water Management. 97(1), pp. 112 122. Kwon, H. Y., S. Grunwald, H.W. Beck, Y. Jung, S.H. Daroub, T.A. Lang and K.T. Morgan. 2010b. Modeling of phosphorus loa ds in sugarcane in a low relief landscape using ontology based simulation. J. Environ. Qual. (39): 1 11. Leduc, D. and J.Goelz 2009. A Height Diameter Curve for Longleaf Pine Plantations in the Gulf Coastal Plain South. J. Appl. For. 33(4):164 170 Lemon M. and O. Kelly. 2009. Laying Second Life foundations: Second chance learners get first life skills. Same places, different spaces. Proceedings ascilite Auckland 2009, 557 565. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. Lindstrom, P., D. Koller, W. Ribarsky, L.F. Hodges, N. Faust, and G.A. Turner. 1996. Real time, continuous level of detail rendering of height fiedls. Proceedings of the 23rd annual conference on Computer graphics and interactive techniques, Anaheim, CA, pp. 109 118, 21 July 1996. Media Grid, 2010, "Create Once, Experience Everywhere" 3D/VR Format Unveiled for Immersive Education, http://mediagrid.org/news/2010 11 _iED_Create_Once_Experience_Everywhere.html
170 Miller, J. A., G. T. Baramidze, A. P. Sheth and P. A. Fishwick. 2004. Investigating ontologies for simulation modeling. Simulation Symposium, 2004 Proceedings, 37th Annual. Nash, J.E. and J.V. Sutcliffee. 1970 River flow forecasting through conceptual models part 1 A discussion of principles. Journal of Hydrology, 10(3), pp. 282 290. Niknam, N. and C. Kemke. 2011. Modeling Shapes and Graphics Concepts in an Ontology. Proceedings of the First Interdisciplina ry Workshop on SHAPES, Karlsruhe, Germany, September 27, 2011. NOAA( National Oceanic and Atmospheric Administration ).2007. NOAA Virtual World. http://www.noaanews.noaa.gov/stories2007/s28 49.htm Noy, N.F., M. Sintek, S. Decker, M. Crubezy, R.W. Fergerson, and M.A. Musen. 2000. Creating semantic web contents with Protege 2000. Journal of IEEE Intelligent Systems, 16(2), pp. 60 71. OLIVE. 2010. http://www.saic.com/products/simulation/olive Oloruntegbe, K.O. and G.M. Alam. 2010. Evaluation of 3d environments and virtual realities in science teaching and learning: The need to go beyond perception referents. Scientific Research and Essays, 5(9):948 954. On http://www.forterrainc.com/index.php/product a services OpenSimulator. 2007. http://opensimulator.org OpenWonderland. 2010. http://openwonderland.org Orland, B., K. Budthimedhee, and J. Uusitalo. 2001. Considering virtual worlds as representations of landscape realities an d as tools for landscape planning. Landscape and Urban Planning. 54:139 148. Palubicki, W., K. Horel, S. Longay, A. Runions, B. Lane, R. Mech, and P. Prusinkiewicz 2009 Self organizing tree models for image synthesis ACM SIGGRAPH 2009 papers, August 03 07, 2009, New Orleans, Louisiana. Parisi, S., J. Baauch, J. Berssenbrugge, and R. Radkowski. 2007. Using Ontology to create 3D Animations for Training Purposes. International Journal of Software Engineering and Its Applications, 1(1), pp. 67 78. Peet, R. K. and D.J. Allard. 1993. Longleaf pine vegetation of the southern Atlantic and eastern Gulf coast: a preliminary classification. Proceedings of the Tall Timbers Fire Ecology Conference. 18: pp. 45 81.
171 Platt, W.J., G.W. Evans, and S.L. Rathbun. 1988. The P opulation Dynamics of a Long Lived Conifer (Pinus palustris). The American Naturalist. 131(4), pp. 491 525. Prada, R. D. Dias, H. Prendinger, and A. Nakasone. 2010. Fostering Agriculture Environmental Awareness. 2010 second International Conference on Game s and Virtual Worlds for Serious Applications. March 25 26, Braga, Portugal. Prusinkiewicz, P. 2002. Art and science for life: Designing and growing virtual plants Development, Evaluation, Production and Use: Proceedings of the XXVI International Horticultural Co Ramasundaram, V., S. Grunwald, A. Mangeot, N.B. Comerford, and C.M. Bliss. 2005. Development of an Encironmental Virtual Field Laboratory. Computers & Education 45(1), pp. 21 34. Rogers, L. 2009. Simulating clinic al experience: Exploring Second Life as a learning tool for nurse education. Same places, different spaces. Proceedings ascilite Auckland 2009, 1121 1129. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. Sanchez Ruiz, A.J., K. U mapathy, and P. Hayes. 2008. Toward Generic, Immersive, and Collaborative Solutions to the Data Interoperability Problem which Target End Users. ONISW '08, October 30, Napa Valley, California. Schott, R.C. 2009. GigaGeology: Virtual Field Trip in a Web2.0 World. GSA Annual Meeting. October 18 21. Portland, OR. Senges, M. and M. Alier. 2009. Virtual worlds as environment for learning communities. In: M.D. Lytras, R. Tennyson, and P.O.D Pablos, Editor, Knowledge Networks: The Social Software Perspective, IGI Global, pp 1 422. SketchUp. 2000. http://www.sketchup.com SimFarm. 1993. Maxis Software Inc. Smalltalk. 1972. http://www.smalltalk.org Speedtree. 2002. http://www.speedtree.com Suman, S., A. Amini, B. Elson, and P. Reynolds. 2010. Design and Development of virtual Learning Environment Using Open Source Virtual World Technology. KCKS, IFIP AICT 324, pp. 379 388. Swanson, B.D. an d D. Kane. 2007. Second Life: Education in a Virtual World. University of California Berkeley, Library Association. http://www.slideshare.net/HVX.Mica/sec ond life education in a virtual world slidecast
172 Virtual Newrological Education Centre. 2007. http://healthinfoisland.blogspot.com/2007/03/virtual neur ological education centre.html Willmott, C.J. 1981.On the validation of models. Physical Geography 2, pp. 184 194. Wood, D. 2009 Experiential learning through 'real world' placements undertaken in 3D 'virtual world' spaces. Same places, different space s. Proceedings ascilite Auckland 2009, 1121 1129. 26th Annual ascilite International Conference, Auckland, 6 9 December 2009. X3D. 2004. http://web3d.org/x3d
173 BIOGRAPHICAL SKETCH Yunchul Jung, hailing from Ulsan, Rep ublic of Korea, finished his schooling from degree from Seoul National University, Seoul. On August 2006, he entered the graduate program at University of Florida. From Augu st 2006 to July 2008, he worked as a Research Assistant at the Department of Agricultural and Biological Engineering, UF, and completed his master's thesis on An Ontology based Approach to Simulation with Application to Citrus Water and Nutrient Managemen t under Dr. Howard W Beck. After completing his Master of Engineering, he started the doctoral program at UF and was awarded a research assistantship under Dr. Howard W Beck in the area of information technology.