<%BANNER%>

Design and Development of Ontology Based 3D Virtual Greenhouses

Permanent Link: http://ufdc.ufl.edu/UFE0021872/00001

Material Information

Title: Design and Development of Ontology Based 3D Virtual Greenhouses
Physical Description: 1 online resource (55 p.)
Language: english
Creator: Appuswamy, Raja
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: 3d, appuswamy, beck, design, eai, esri, graphics, greenhouse, greenhouseontology, howard, java, lyra, object, ontology, owl, plantontology, raja, sai, scripting, sketchup, swing, taxonomy, virtual, virtualreality, vr, vrml, walkthrough, x3d, xj3d
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: With the growing popularity and widespread appreciation for the design and construction of greenhouses, several attempts have been made over the past few years in developing web based virtual greenhouses. The approach taken towards the design of such greenhouse can be classified into two types. The first type focuses on delivering knowledge to the users. Visualization is either not a part of such a system or is minimal. The second type is centered on developing a visualization of the greenhouse so that the users can get a feel of the real world peer. Such a visualization system stands alone as a separate entity and is not linked with any meta information. This research focuses on developing a 3D virtual reality based virtual greenhouse. The 3D model is coupled with information sources such as text based descriptions and relevant images by using a single data store (Lyra OMS) to store both the 3D model and meta information. Using Lyra made it possible to link objects in the 3D scene with domain ontologies and thus build a rich 3D virtual environment in which users can create new greenhouses, save and load previously saved greenhouse instances from the database, walk through the 3D greenhouse model, inspect 3D objects and view all relevant information. They are able to tag objects with names and group multiple 3D objects with the same name. A search functionality that redirects the user to his/her point of interest within the 3D space has also been implemented. An energy loss estimation engine operates on the user specified properties of the greenhouse and produces statistics of energy loss per structural component. Users can compare the energy loss characteristics of multiple greenhouse configurations and refine their design. An educational simulation that shows the status of an electric heater with the internal temperature has been implemented to illustrate the capability of extending this system to support simulations.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Raja Appuswamy.
Thesis: Thesis (M.S.)--University of Florida, 2007.
Local: Adviser: Beck, Howard W.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021872:00001

Permanent Link: http://ufdc.ufl.edu/UFE0021872/00001

Material Information

Title: Design and Development of Ontology Based 3D Virtual Greenhouses
Physical Description: 1 online resource (55 p.)
Language: english
Creator: Appuswamy, Raja
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2007

Subjects

Subjects / Keywords: 3d, appuswamy, beck, design, eai, esri, graphics, greenhouse, greenhouseontology, howard, java, lyra, object, ontology, owl, plantontology, raja, sai, scripting, sketchup, swing, taxonomy, virtual, virtualreality, vr, vrml, walkthrough, x3d, xj3d
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: With the growing popularity and widespread appreciation for the design and construction of greenhouses, several attempts have been made over the past few years in developing web based virtual greenhouses. The approach taken towards the design of such greenhouse can be classified into two types. The first type focuses on delivering knowledge to the users. Visualization is either not a part of such a system or is minimal. The second type is centered on developing a visualization of the greenhouse so that the users can get a feel of the real world peer. Such a visualization system stands alone as a separate entity and is not linked with any meta information. This research focuses on developing a 3D virtual reality based virtual greenhouse. The 3D model is coupled with information sources such as text based descriptions and relevant images by using a single data store (Lyra OMS) to store both the 3D model and meta information. Using Lyra made it possible to link objects in the 3D scene with domain ontologies and thus build a rich 3D virtual environment in which users can create new greenhouses, save and load previously saved greenhouse instances from the database, walk through the 3D greenhouse model, inspect 3D objects and view all relevant information. They are able to tag objects with names and group multiple 3D objects with the same name. A search functionality that redirects the user to his/her point of interest within the 3D space has also been implemented. An energy loss estimation engine operates on the user specified properties of the greenhouse and produces statistics of energy loss per structural component. Users can compare the energy loss characteristics of multiple greenhouse configurations and refine their design. An educational simulation that shows the status of an electric heater with the internal temperature has been implemented to illustrate the capability of extending this system to support simulations.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Raja Appuswamy.
Thesis: Thesis (M.S.)--University of Florida, 2007.
Local: Adviser: Beck, Howard W.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2007
System ID: UFE0021872:00001


This item has the following downloads:


Full Text





DESIGN AND DEVELOPMENT OF ONTOLOGY-BASED 3D VIRTUAL GREENHOUSES


By

RAJA APPUSWAMY
















A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2007

































2007 Raja Appuswamy



























To my mom, dad and friends









ACKNOWLEDGMENTS

I thank Dr.Howard Beck, my advisor and chair, for all the support he has offered me

during the course of my studies as a graduate student. His style and approach to problems

profoundly influenced me and helped me hone my technical knowledge.

This research would have been impossible without the support offered by Dr.Ray Bucklin.

I would like to thank him profusely for guiding me and providing me with the right resources at

the right time.

I am indebted to Professor James Oliverio as my experience working with him on ISAS

greatly influenced the design of the 3D virtual greenhouse.

I would like to thank my parents for I would have never made it this far without their

blessings and last but not the least, I would like to thank all my friends whose constant

encouragement and support made everything possible.









TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ........... ................. ...............................................................................4

L IST O F T A B L E S ................................................................................................ . 7

LIST OF FIGURES .................................. .. ..... ..... ................. .8

A B S T R A C T ................................ .................. .......................... ................ .. 9

CHAPTER

1 IN TR OD U CTION ......................................................... ................. .. ........ 11

2 L IT E R A TU R E R E V IE W ......................................................................... ........................ 13

2.1 UMD, UW- Madison Greenhouse Virtual Tour.................................... ...............13
2.2 University Of Florida Virtual Field Day ....................................................................14
2.3 The Greenhouse Simulator by University of Vermont...............................................14
2 .4 V virtual G row er............ ... ..... .............................................................. ............ 16
2.5 Visualization of Environment by VRML ...................................................................17
2 .6 E v alu atio n ................................................................................ 18

3 TO O L S A N D TE CH N IQ U E S........................................................................ ..................20

3 .1 Jav a Sw in g ............................................................................... 2 0
3 .2 V R M L ................... ...................2...................1..........
3 .4 X j3 D ..........................................................................2 4
3.5 Lyra Ontology Management System .................................................................. 25
3.6 A dressing the L im stations ..................................................................................... ...... 30

4 ONTOLOGY BASED 3D VIRTUAL GREENHOUSE ................................................33

4 .1 D o m ain O n to lo g ies .................................................................................................. 3 3
4.2 Database Design for 3D Objects ................................ ........................ ......35
4 .3 U se r In te rfa c e ........................................................................................................... 3 6
4 .3 .1 M eth o d o lo g y ................................................................................ 3 6
4 .4 M en to r M o d e ........................................................................................................... 3 7
4 .4 .1 M eth o d o lo g y ................................................................................ 3 8
4 .5 In sp e ct M o d e ........................................................................................................... 3 8
4 .5 .1 M eth o d o lo g y ................................................................................ 4 0
4 .6 S e a rc h in g ................................................................................................................. 4 2
4 .6 .1 M eth o d o lo g y ................................................................................ 4 2
4 .7 E n erg y E stim atio n ................................................................................................... 4 4
4 .8 C o m p ariso n M o d e ................................................................................................... 4 6
4.9 Simulation mode ................................. .............................. ........ 47









4.10 M methodology ....................................................... ... ........ .. ............ 47

5 CONCLUSIONS AND FUTURE WORK....................................................................... 49

5.1 O ntology based search .................................................... ... .. ........ .... 49
5.2 N natural language Support ........................................... .................. ............... 50
5.3 Geo referencing the 3D m odel............................... ................. ... ...................... 51
5.4 Exporting the 3D model to other formats............. ........................... ...............52

L IST O F R E F E R E N C E S ......... ................. ...............................................................................53

B IO G R A PH IC A L SK ETCH .......................................................................................... 55










































6










LIST OF TABLES


Table


5 -1 U v alu e s ...........................................................................................4 7


page









LIST OF FIGURES

Figure page

2-1 Web page showing the 360 degree Quicktime virtual tour .............. ...........................15

2-2 G reenhou se sim ulation ............................................................................. .................... 16

2-3 Im ages of V virtual G row er ..................................................................... ....................... 17

2-4 VRML Greenhouse developed at Chiba University ........... ...................................19

3-1 D om ain ontology for Com puters ............................................... ............................ 30

4-1 G greenhouse dom ain ontology.. .............................. ... .......................................33

4-2 T om ato P lant ontology ............................................................................ .....................35

4-3 Swing based UI for setting the properties of the greenhouse. ................. ...............37

4-4 The 3D greenhouse rendered in Xj3D. ........................................ ......................... 39

4-5 Textual Information Pane containing comments, user defined names and viewpoints.....40

4-6 Image of an electric heater being added to the 3D object...............................................41

4-7 Aggregating shape nodes under a common user defined name ......................................43

4-8 Search dem onstration ............................................................. .............. .. 45

4-9 Simulation with the heater showing the internal temperature in the greenhouse ............48

6-1 Images showing geo-referenced Rinker Hall ........................................ ...............52









Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master Of Science

DESIGN AND DEVELOPMENT OF ONTOLOGY BASED 3D VIRTUAL GREENHOUSES

By

Raja Appuswamy

December 2007

Chair: Dr. Howard Beck
Major: Agricultural and Biological Engineering

With the growing popularity and widespread appreciation for the design and construction

of greenhouses, several attempts have been made over the past few years in developing web

based virtual greenhouses. The approach taken towards the design of such greenhouse can be

classified into two types. The first type focuses on delivering knowledge to the users.

Visualization is either not a part of such a system or is minimal. The second type is centered on

developing a visualization of the greenhouse so that the users can get a feel of the real world

peer. Such a visualization system stands alone as a separate entity and is not linked with any

meta information.

This research focuses on developing a 3D virtual reality based virtual greenhouse. The 3D

model is coupled with information sources such as text based descriptions and relevant images

by using a single data store (Lyra OMS) to store both the 3D model and meta information. Using

Lyra made it possible to link objects in the 3D scene with domain ontologies and thus build a

rich 3D virtual environment in which users can create new greenhouses, save and load

previously saved greenhouse instances from the database, walk through the 3D greenhouse

model, inspect 3D objects and view all relevant information. They are able to tag objects with









names and group multiple 3D objects with the same name. A search functionality that redirects

the user to his/her point of interest within the 3D space has also been implemented.

An energy loss estimation engine operates on the user specified properties of the

greenhouse and produces statistics of energy loss per structural component. Users can compare

the energy loss characteristics of multiple greenhouse configurations and refine their design. An

educational simulation that shows the status of an electric heater with the internal temperature

has been implemented to illustrate the capability of extending this system to support simulations.









CHAPTER 1
INTRODUCTION

Planning, designing and constructing a greenhouse has become an activity of interest to a

wide range of people from hobbyists to horticultural experts. To encourage people to develop

greenhouses and to educate people worldwide about developing such structures, the concept of

developing a "virtual greenhouse" has become very popular. These virtual greenhouses are web

based electronic representations of their real world peers. The purpose of such greenhouses is to

provide a repository of information which users can utilize to make quality decisions about

constructing their own greenhouse or selecting which plants to grow in what types of

greenhouses. Some virtual greenhouses link to diverse multimedia information sources like

images, educational videos and Flash animations. The user navigates through the model and

follows the links to his/her area of interest.

With the advent of 3D on the Internet, one of the ways to model a real world greenhouse is

to construct a model in 3D. Several virtual reality (VR) environments exist that import such 3D

models and render them. Using these VR environments, the users are able to walk through the

virtual greenhouse and look at plants or other components at a various levels of detail.

Traditionally 3D virtual environments utilize files (VRML, X3D) that describe the

geometry of 3D objects but contain very little additional knowledge of the objects in the VR

world. There are many reasons to enhance VR environments with better database support. In a

VR world, we need to know not only what an object looks like, but also what the object is, what

its properties and characteristics are, how it behaves, and how it relates to other objects. We also

need to be able to interact with and refer to objects by pointing or by voice commands.

In our approach we utilize an ontology management system, a database management

system based on ontologies, to provide semantic descriptions of objects in the VR environment.









In fact the VR environment can be considered a projecting of objects described by the ontology

into a 3D visualization environment. The ontology provides a way of describing object properties

and behaviors, describing object taxonomies, and, using ontology reasoners, automatically

classifying and clustering objects into categories. The ontology acts as a dictionary that defines

the meaning of concepts, and provides support for natural language references to objects.

This thesis focuses on studying the practicability of a virtual reality based 3D environment

which is augmented with meta information. The purpose of this research is to build a unified VR

environment which offers its users the capability to walk through the 3D model and also inspect,

add, modify and search for meta information on the fly. The meta information provided by the

user is then used to relate objects in the 3D scene with domain ontologies. Tagging 3D shapes

with ontologies builds up a taxonomy that can be navigated by the users to find relevant

information. In addition, coupling semantic content with the 3D shapes also makes future

extensions possible, such as using a reasoner to replace a text based search with semantic search.









CHAPTER 2
LITERATURE REVIEW

University of Minnesota Duluth (UMD), University Of Wisconsin Madison Greenhouse
Virtual Tour

UMD virtual tour [1] was developed to expose the users to the different groups of plants in

the UMD greenhouse and to help the users understand the way the plants and benches are

organized. The tour was simulated by a collection of web pages, each providing in-depth

information about the greenhouses. The web site contains an index page that provided the

visitors with general information and contains links to the two UMD greenhouses, the Upper

Greenhouse and the Lower Greenhouse. Users could navigate to the greenhouse of their interest

by following the link. Each virtual greenhouse is modeled as a web page using HTML. The web

page contains an image that shows the floor plan of the greenhouse. The floor plan describes the

location of plants, benches and other fixtures within the greenhouse and is image mapped to

internal links within the web page. Users can click on individual parts of the floor plan and jump

to the content of their interest, or they could simply scroll down and view the entire content.

Information about each plant, bench and fixture comes in two forms: a paragraph of textual

information that describes the relevant information and photographs pertinent to that component.

The design of the UW- Madison greenhouse is similar to that of UMD. Use of hyperlinks

to model the relationship between rooms in a greenhouse was the adopted technique. The index

page contained hyperlinks to all rooms within the greenhouse. Each room was modeled as a web

page that contained its own set of hyperlinks that linked to individual plants within the

greenhouse. Each individual plant had a web page that provided detailed information about the

plant and relevant pictures.









University Of Florida Virtual Field Day

The Virtual Field Day project [2] undertaken by the University Of Florida aims at

providing the users access to a repository of information about the research greenhouses at

Gainesville and Suwannee. The Virtual Field Day project provides its users with an "Interactive

Map" which is a Flash object embedded in a web page. The users can view the greenhouses in

the map. The map assists the users in navigating around the locations in the Suwannee or

Gainesville Virtual Field Day. Selecting a location takes them to that location's multimedia

resources. In addition to the text based and image based information, the site's multimedia

resources include instructional videos, publications and research literature pertinent to the

technologies used in the greenhouse. In addition, each greenhouse has a Quicktime VR tour

which enables the users to interact with the 360 degree panoramas. The tour plays in Quicktime

player and has tags on specific areas of interest. Navigation through the tour is limited to a

stationary 360 degree view and zooms in/out. Clicking on tags takes the user to a web page that

contains a listing of related videos, publications and other instructional material. Separate 360

degree views show the users how the greenhouse looks in the pre-plant season and during the

main season.

In both the Virtual Field Day project [2] and the greenhouses by UMD [1] and UW,

photographic images serve as the primary means by which the users can visualize the

greenhouse. All the information is static and the users cannot simulate changes in the greenhouse

environment based on properties like climate or structure. User interactivity is limited to the

traditional web based clicking and hyper linking.

The Greenhouse Simulator by University of Vermont

The Greenhouse simulator [3] is a part of a world wide greenhouse education initiative [4]

undertaken by The University of Vermont, The University of Florida, University of Arizona















Main Season 1
Starting the Crop







Srnlainstructional videos related to seeding, transplants etc on the left Virtual Field Day(11)
Nutri ront aSoltiotn DOflivery
N,,* t Conn s 1t "* (4:54)


ier the u ec ue e e ee ue ee )
Pemaking st M anentgeographic, and seasonal choices. The simulator incorporates
climate co itio a 6 Inde grthe aphidroic Hdy roporic chre house
M- -initrinq Comnn,- PIt 434)
(5:19)

Figure 2-1 Web page showing the 360 degree Quicktime virtual tour on the right and
instructional videos related to seeding, transplants etc on the left Virtual Field Day
Project at University Of Florida [2].

and Ohio State. The outcomes of this project include a greenhouse simulator, a digital repository

and educational materials. The digital repository contains a searchable collection of images and

videos. The simulator is a Flash object that allows users to model the greenhouse environment by

making structural, environmental, geographic, and seasonal choices. The simulator incorporates

user-selected information from its database of greenhouse designs, operation, and geographic

climate conditions, and graphically displays dynamic changes in greenhouse environments,

including moist air properties. The animation allows learners to simulate changes in the

greenhouse-plant environment based on climate, structure, and environmental control choices.

The interactive greenhouse environment simulator was developed by integrating

mathematical models and an animation interface, which was created using Flash MX Pro

2004.The greenhouse mathematical model, which is based on the energy and mass balance of the

greenhouse system, utilizes a series of differential equations. The solution provides the dynamic

response of the greenhouse climate conditions to the outside climate conditions and for a

particular greenhouse design. The design incorporates user-selected inputs for climate, structure,










glazing, and environmental control systems. Each simulation demonstrates the response of a

greenhouse system design over a 28-hour period.

The simulator is a stand-alone piece of software and is not linked to any other information

sources. The user interface limits the interaction to the user selection of various parameters.

Further information about components inside the greenhouse cannot be obtained from the

greenhouse model in the simulation.


I I]F L i. 'I. I%-r-.l1r.i j-t Er
Pro ** .. .. G r o L Ildo, lo. t i a. die t I. 1

| on sru r. b ,*. ca |l L_ __de'_ 1O _s_ _r ef o ga si s n o c t-' b4Sock__ _________
d- r
LKak- 3. 4




..., -..... .



h.................irt u a l....." .. ...............
database gives a year's worth of typical weather for the city nearest to the greenhouse location
..including factors such as temperature, sunlight and cloud cover for each hour of the day.
H II .1 1

Figure 2-2 Greenhouse simulation shown plotting values of solar radiation, air temperature,
humidity etc over a period of 28 hours based on user specified inputs. [3]

Virtual Grower

Virtual Grower [5] is a PC based stand-alone program developed by the ARS Greenhouse

Production Research Group at Toledo, Ohio. It is a decision support tool for greenhouse growers

that enable the users to estimate the energy requirements of a greenhouse. To estimate energy

requirements and costs using Virtual Grower, users input the dimensions of their greenhouse and

its construction materials, such as poured-concrete floor, glass sides and roof, or concrete-block

walls. They also choose design features, such as roof shape and orientation to the sun. A historic

database gives a year's worth of typical weather for the city nearest to the greenhouse location

including factors such as temperature, sunlight and cloud cover for each hour of the day.











Growers choose a heating schedule and set the temperatures they want to maintain during day


and night, or for each hour. The program then calculates per-square-foot heating costs by the


month or year. The users can save their heating schedule, greenhouse configuration or export


their results to other formats.



Name DI-v H rensns and ihenta on
NMaumber of Spans
FMate rial s Length f Span r5 lee
Floor '.ed onee Wdth olSpan leet -
Sidng Maleral RoolShape Trangular v -
North [las v KneewallHeight leet
South Gla Height on the Edge feet
East [ G as Helghtin the Middle leet

Knee Wal Materal O'rientation h So
North Concrete Block FuelType
Southh Concr etBlock v FuelType FueOliri ~
East ConcreteBlock Fuel Pce 12 pe gallon
West ConcreteBlock HeaterEiffciency 5 percent ,




Height a
Height at Edge 19Eddle J d | |

A Numbe of Sp ans


Figure 2-3 Images of Virtual Grower[5] A) User setting various parameters of the greenhouse B)
Virtual Grower reporting the estimated heating costs

The visualization of the greenhouse constructed by the users is limited to static line


diagrams that depict the structure of the greenhouse as shown above. The primary purpose of this


software is cost estimation and hence visualization is not a component. User interaction is also


limited to selecting the various parameters.


Visualization of Environment by VRML

Honjo. T and Lim. E [6] designed a visualization system for greenhouse that enables


virtual experience in a planned environment was developed using VRML. Their research focuses


on walk through simulations as a tool for understanding the design of landscapes and


greenhouses. 3D plants and objects with a high polygon count generally take a very long time to


render and require expensive, high power hardware for a rich walk through experience. To avoid









this problem, a plant was expressed by using two textured planes instead of thousands of

polygons. Computer generated images of plants at several growth stages are made by using

AMAP (Atelier de Modelisation de Architecture de Plants) [7], which is the system developed

by CIRAD ( Center Internationale Recherche Agricultural Development). The plant images then

serve as texture for the two planes thereby rendering the plants with minimal overhead. The

greenhouse structure itself was made out of simple VRML constructs. The VRML virtual

greenhouse enabled walk through. The 3D model was also posted on the web so that users could

use a web browser to view the VRML world.

Having the user walk through the 3D model provides a better understanding of various

parts of the model since its equivalent to his/her walking through the real world greenhouse. This

system focuses only on the visualization aspect. It does not relate 3D models with relevant

information or other multimedia sources like images or educational videos. The 3D scene is

static since it comes off a VRML file. The user cannot make changes to the 3D scene at runtime

or add information to certain components. For instance a user who has an image of a corn plant

will not be able to add it to the collection of existing images.

Evaluation

Research literature indicates the need for a system that merges the pros of both 3D based

virtual greenhouses and other HTML based models. While the 3D greenhouse provided a better

user interaction and navigation capability, it was not associated with multimedia information

sources. Compared to this, the traditional web based greenhouses had images, videos and text

based information sources. The disadvantage with the HTML based virtual greenhouses was that

visualization of the greenhouse was limited to photographs and walking through the greenhouse

to foster a better understanding of the real world greenhouse was not possible with plain images.











'- ~* c -- : -- _



















Figure 2-4 VRML Greenhouse developed at Chiba University [6]











































19


"-









CHAPTER 3
TOOLS AND TECHNIQUES

The following design goals were formulated:

* Build a 3D model of a greenhouse through which the users can walk.

* Provide a mechanism by which plants and other components in the 3D model can be
annotated with meta-information and also inspected to view existing information sources.

* Store the model and meta-information in an ontology management system that can be
accessed remotely.

* Implement a search functionality so that the users can search and be redirected to see the
objects of their choice.

* Develop an educational simulation involving thermal characteristics of the greenhouse.
The simulation will provide visual cues that depict switching a heater on and off based on
heat loss from the greenhouse. This adds dynamic behavior to the 3D greenhouse
environment.

To meet all the these goals, the following tools and techniques were employed.

Java Swing

Java Swing [22] is part of the Java Foundation Classes, a group of classes that encompass

features for building GUIs and adding rich graphics functionality and interactivity to graphics

applications. The other components that form a part of JFC include

* PLAF (Pluggable Look And Feel)

* Accessibility API to enable assistive technologies like screen readers

* Java 2D API to help developers enrich their applications with 2D graphics and text,

* Internationalization to enable the development of programs that can interact with the users
in their own local language and cultural conventions

* Drag and Drop.

Prior to Swing, Abstract Window Toolkit (AWT) was used to construct user interfaces. AWT

had several problems associated with it. It was not sufficient for complex user interfaces, it was

not portable, it heavily relied on the runtime platform to render many of its user interface









components and had a poor event model. Swing was developed on top of AWT to solve many of

the problems associated with it. Swing provides pluggable look and feel. All components in

Swing are lightweight in the sense that they do not depend on native runtime environment except

for a few top level containers. Several new components such as sliders and progress bars are

added to the list of existing AWT components. This research uses Java Swing to build the User

Interface.

VRML

VRML (the Virtual Reality Modeling Language) [8] is an international standard for

describing 3-D shapes and scenery. VRML files describe a 3D world. Visualizing the 3D world

can be done by viewing the VRML file in a VRML browser. Since many of the VRML browsers

are available as web browser plug-ins, 3D information can be easily published on the web using

VRML.

The VRML file contains a node hierarchy which reflects the way in which individual

shapes in the 3D scene are grouped together to form composite objects. Grouping nodes in the

VRML files serve to group together their children. Transform nodes, in addition to grouping a

set of nodes also add a translation, rotation or scale to all of its child nodes. Shape nodes contain

information about the appearance and geometry of each shape in the 3D scene and are grouped

within grouping nodes. The appearance of the shape node is usually provided by an Appearance

node that controls several properties like coloring aspects, textures, transparency and opacity.

The geometry of a shape node can be described by making it one of the predefined types like a

BOX or by using an IndexedFaceSet node. An IndexedFaceSet node lists the coordinates of the

points that form the shape and groups up points that form an individual face of the shape, one

group per face, thereby building up the structure. Nodes in a VRML file are named by using the

DEF construct. Prefixing a node declaration with a user defined name assigns the user defined









name to the node and makes it possible to refer to that node anywhere in the file using that name.

USE construct makes use of such "DEFed" nodes to reuse existing node declarations. For

example, in the following node hierarchy given below, an appearance node and a geometry node

get DEFed with user defined names and they are reused further down the hierarchy to control the

appearance and geometry of another shape.

Shape {

appearance DEF myAppearance Appearance {



}

geometry DEF myGeo geometry {



}

}


Transform {

translation 0 0 0

Shape{

appearance USE myAppearance

geometry USE myGeometry

}

}

The 3D scene built by the VRML nodes is static. To add animations or to make it

interactive based on the user input, the VRML nodes must be connected to each other and

messages must be passed between them. For this purpose, most of the nodes have special fields

called eventIns and eventOuts. EventIn fields are like receivers, which receive messages called









events from the outside and take them in to be processed. EventOut fields are transmitters, which

send events from the node to the outside. Nodes are connected to each other by attaching these

fields using ROUTEs. These are like pipes that channel events from an eventOut into an eventIn.

Many eventIns can be connected to one eventOut to make one event cause many things to

happen. This is called fan-out. Fan-in is also allowed, where two or more eventOuts feed into

one eventIn.

VRML also contains some special nodes. These nodes act like sensors in that they sense

user interaction with the scene and react to it by throwing events. One such node that provides

interactivity in the 3D greenhouse is the Touch Sensor node. This node detects mouse interaction

with its sibling geometry. Just like other nodes, TouchSensor also has some eventIn and

eventOut fields. isOver, isActive and touchTime are the eventOuts in TouchSensor. When the

mouse is moved over the geometry, an "isOver TRUE" event is generated, and when it moves

off, an "isOver FALSE" is sent. When any of the mouse buttons is clicked while isOver is

TRUE, two events are generated, namely, an "isActive TRUE" event and an the "touchTime

" ( where currentTime indicates the time when the user clicked on the shape )

event is generated. To illustrate the process of creating animations in a VRML scene, consider a

scene that has a touch sensor node and a Sound node. The Sound node is used in VRML for

generating 3D sounds. When the touchTime eventOut of the TouchSensor is connected to the

beginTime eventOut of the Sound node using a ROUTE, the sound file specified in the sound

node is played when the user clicks on the shape that the TouchSensor monitors.

VRML files can be hand coded or generated from third party applications. The 3D models

for the greenhouse are fairly complex. Hence, a third party software modeling tool SketchUp,

was used to build the 3D model for each of 6 different types of greenhouses. The models were









then exported to generate VRML files.


Xj3D

VRML as such provides a language by which modelers can describe a 3D scene. VRML

by itself cannot display a 3D scene without the software to render the 3D scene. To render the

environment, a VRML browser is required. Xj3D [11] is a project of the Web3D Consortium

focused on creating a toolkit for VRML97. X3D content is written completely in Java and

provides a VRML browser that can be used to render a 3D scene. We can use Xj3D to render the

VRML scene from a VRML file or use its API to build up the scenegraph directly by

constructing nodes one at a time and adding them to the scene. Xj3D also has a parser

implementation that can be used to parse a VRML file and generate an in memory representation

of the VRML node hierarchy.

Interactivity and animations can be added to the VRML file by two mechanisms. The first

one is scripting within VRML. Scripting languages like ECMA Script [10], JavaScript can be

placed inside special type of nodes called Script Nodes. These scripts process on internal events

and change the nodes in the scenegraph. The other way is to use EAI (External Authoring

Interface) [1 1].The EAI enables control of the contents of a VRML browser window embedded

in a web page from a Java (tm) applet or such an external source on the same page. The

difference between EAI and internal Script node is that EAI offers a generalized method to

access nodes and events in the VRML browser's scene graph by connecting to a VRML browser

plug-in through a browser plug-in architecture like Live Connect or ActiveX/COM. Internal

Script Nodes have node access through definition within a VRML Script Node within a VRML

scene graph. The EAI supports events and fields on a dynamic basis, while the Script Nodes'

definitions consist of predefined fields and events. Xj3D merges the EAI and scripting paradigm









and provides a single unified interface called SAI (Scene Access Interface). It provides a

compatible way to run and modify X3D and VRML scenes for many X3D implementations.

The virtual greenhouse uses the Xj3D API for two purposes. Xj3D Parser API is used to

parse the VRML file. The parser operates based on an evening paradigm. The parser expects the

caller to register a content handler with it. The content handler contains a list of functions that are

called back by the parser at critical points such as document entry/exit, node entry/exit and field

entry/exit. For instance, when the parser hits the beginning of a node, it calls the function

startNode on the content handler. The content handler contains a set of stacks that it uses to pass

information between calls to the functions. For instance, when the start Node function is called,

the content handler creates a new hash map keyed on field names to store the field values for this

node and pushes it onto the top of the stack. When field data is encountered by the parser, it calls

the fieldValue method which creates a new entry in the hash map. When a node definition ends,

the parser calls the endNode method which removes the hashmap from the top of the stack. Thus,

using the parser API and a custom content handler implementation, an in memory representation

of the VRML file can be generated.

The second place where Xj3D API is used is to provide an environment for 3D

visualization. Xj3D provides a VRML browser implementation that is used to host the 3D

greenhouse. VRML nodes are constructed based on the shape information present in the database

and the nodes are added to the scene graph which renders itself once constructed completely.

Thus, using SAI, the application interfaces with the database and updates the scene at run time

during the simulation.

Lyra Ontology Management System

In philosophy, ontology is a theory about the nature of existence, of what types of things

exists; ontology as a discipline studies such theories. The most typical kind of ontology has









taxonomy and a set of inference rules. The taxonomy defines classes of objects and relations

among them. The most basic concepts in a domain form the roots of various taxonomic trees as

classes. A concept more specific than the general high level concept is considered to be a

subclass. The generic concept becomes the superclass of the sub class. Members of a class are

called instances. Instances can be thought of as corresponding to individuals in the real world.

Properties assert general facts about the members of classes and specific facts about individuals.

Let us consider the plant ontology given below.

Kingdom: Plantae

Division: Magnoliophyta

Class: Liliopsida

Order: Cyperales

Family: Poaceae

Genus: Zea

Species: Zea mays

Kingdom, Division, Class, Order, Family and Genus are all classes representing concepts

in this ontology forming a taxonomic hierarchy for classification of plants. Plantae,

Magniliophyta etc are instances of the corresponding classes. The kingdom Plantae indicates that

corn is a plant and probably photosynthesizes using chlorophyll. The division Magnoliophyta

indicates that it is a flowering plant. The class Liliopsida indicates that corn is a monocot-that

is, it has one cotyledon (the first leaf formed on a seedling) and leaves with parallel veins, and its

flower parts occur in multiples of three (three or six petals, three or six stamens, and so on). The

order Cyperales indicates that the seeds store starch and the flowers lack petals and sepals. The

family Poaceae indicates that it has a unique type of inflorescence, round stems, and a distinctive

internal anatomy and so on, down to the species level.









An ontology management system (OMS) is a database management system that utilizes

formal ontology languages as the basis for modeling and manipulating data. The OMS provides

facilities typically available in database management systems including maintenance of database

integrity, physical storage management, security, transactions, backup/recovery, and data

manipulation. Using the OMS's data definition language, onotologies can be constructed, and

using traditional data storage management techniques, the ontology can be stored and retrieved.

Physical storage management is particularly important because the VR environment requires

management of very large numbers of objects. Dynamic scene generation requires that just the

right objects to be loaded into memory at the appropriate time to expand a scene or increase level

of detail in object presentation. This research uses Lyra [12], an Ontology Management System

as a core database management facility. Lyra stores data in the form of objects. An object in Lyra

is variable length block of bytes. Each object can be identified using either an integer Object

Identifier (OID) or a string Object Identifier (SOID). Lyra creates hash based indices on both

OIDs and SOIDs to enable rapid lookup of objects.

Lyra's data definition language represents all generic concepts as classes and individual

members as instances. Attribute is the term used to represent a property in Lyra. Attributes are

relationship between objects. They have a domain and a range and they are directional, meaning

that they connect the source objects in the domain to the target objects in the range. For instance,

an attribute "made from wood" could be used to connect a door in a greenhouse with a type of

wood. Thus the domain would be class door and the range would be class wood. Attributes can

be used in either classes or in instances. When used with classes, attributes are used to restrict the

set of values that the class instances can have for that attribute. Such a restriction imposed on the

attribute values is called an attribute restriction. Cardinality is an example of an attribute









restriction. Assigning a minimum and maximum cardinality restriction to an attribute enforces

that the attribute appears at least minimum times and no more than maximum times in the class

instances. For example, when an attribute restriction is created for the door class that specifies

that the cardinality of the attribute "made from wood" to be one, all instances of the door class

must definitely specify exactly one value for the attribute "made from wood".

When an attribute is used in the context of an instance, it is called an attribute instance.

attribute instance describe properties of an instance. For instance, the attribute "made from

wood" could be used in an instance "doorl" to specify that door is made from pine. Thus the

attribute instance of type "made from wood" relates door with the wood instance pine. Lyra

provides two ways for storing attribute instances in the database, namely, storing each attribute

instance as an object or grouping together a set of attribute instances and embedding the group of

values within an object rather than making individual objects. Attribute vector is the term used in

Lyra to represent the grouping of attribute instances. Each attribute vector has a name and

maintains a hashtable that stores all the attribute instances as a set.

Lyra supports several data types from simple ones like integers and strings to more

complex ones like images and videos. The values that attributes take in an instance are stored in

the Lyra as "Data" objects. A Data object is actually a generic object type that is used in Lyra to

store all types of data in binary format. Values are coerced and type casted to their corresponding

data types as and when necessary.

For example, consider the problem of representing a 3D point in Lyra. The coordinate of a

point in 3D space can be represented using 3 values: x, y and z. Each of the 3 values is of

floating point type. To model this in the form of Lyra objects, three attributes representing the x,

y and z properties of a point are created. Let us call the attribute objects aX, aY and aZ and their









OIDs xOID, yOID, zOID. The x coordinate value will be wrapped in a data object. An attribute

instance will be created and associated with aX using its oid xOID and the data object that wraps

the x value will be added to the attribute instance. Similarly, attribute instances will be created,

one for every other coordinate value. An instance is created with the name "Point" and all the

three attribute instances are added to this instance.

Extending this approach to model a complex shape is relatively simple. Consider a box

shape that has 8 end points. Each point is represented using the same methodology as above

with the exception that instead of creating an instance for each point, an attribute vector called

"Point" is created and attribute instances for each of x, y and z coordinates is added to it. An

instance is then created with the name "Box" and the attribute vector is added to it.

Lyra being an OMS can be used to store domain ontologies. Having both the 3D model

and domain ontologies stored in a single data store makes it possible to tag shapes in the 3D

model with semantic information. For instance, consider a 3D scene where a box is used to

represent a computer. Assume that we have a domain ontology stored in Lyra that provides

taxonomy of computer types as shown in Figure 3.1.

If we specify that the box in the 3D scene represents a Desktop Computer instance, then the 3D

scene automatically gets enriched with the domain ontology about computers. Since the box is a

desktop computer, it can be inferred that it is automatically a digital computer. We could use

another ontology that lists the average price of digital computers to get a rough estimate of the

price of the desktop computer. Thus, using an OMS helps us combine the 3D visualization aspect

with the semantic aspect to produce an ontology based 3D visualization environment.

Connecting to the Lyra database is done using RMI (Remote Method Invocation). RMI

provides for remote communication between programs written in Java. RMI allows java objects










running on one system to connect to and invoke methods on an object running in another

machine. A typical RMI server creates the remote objects, makes them available for access and

waits for clients to invoke methods on the objects. An RMI client obtains references to the

remote object and invokes methods on the object. RMI takes care of the communication between

the client and the server. An RMI server registers or binds its remote object usually with the

simple naming facility provided by RMI called the RMI registry. The client looks up the remote

object by its name from the RMI registry in the server and invokes a method on the object. Lyra

uses RMI to enable the users to connect to a database from anywhere.

Computer



Analog Digital Node Number Server


I I I I I
Computer Computer Cruncher


File Server Mainframe Minim Computer Personal Workstation
Computer


Desktop
Computer


Portable
Computer


Figure 3-1 Domain ontology for computers

Thus, using Lyra enables us to add structure to the data by organizing the data with well

defined domain ontologies and makes remote access possible through RMI.

Addressing the Limitations

In the design of landscapes and greenhouses, walk through simulation in a 3-D space is a

useful educational tool, especially for understanding design and for selection of alternative

designs. Walk through simulations serve to provide a rich user experience as highlighted in [6].









Storing the entire model in a database and operating off the database provides three

advantages. The first advantage is that it gives complete control over building the 3D scene. The

3D scene can be modified as it is loaded or after it is loaded depending on the situation. The

entire model or parts of it could be pulled up at runtime depending on the necessity. For instance,

we could render the 3D greenhouse without the plants and heaters or with them depending on

whether the user is interested in visualizing the structure of the greenhouse or inspecting the

components inside the greenhouse. If a static VRML file is used, unless we provide a custom

VRML browser implementation, all modifications to a scene can take place only after the

scenegraph is constructed and hence partially loading a scene would not be possible.

The second advantage is that it simplifies the process of tracking changes to the 3D model.

Changes made to the 3D model can be saved by updating the existing model in the database.

Updating a 3D model does not present any problems as it would in the case of a static VRML

file. Xj3D does not directly support exporting a 3D scene at runtime to a VRML file and even if

we altered the SAI implementation to support this, we would have to write it out to a new VRML

file. This presents several problems since the user could make multiple changes and saves over a

period of time and each save should either save the entire scene, which impairs user interaction

due to disk write delays, or maintain some way of incrementally saving the changes to the new

file. As a result, there would be multiple copies of the same VRML file. These problems can be

avoided using a database because we can change the 3D model and make updates to the database

to reflect the changes upon saving and thus having a single instance of the 3D model. The third

and the biggest gain from using an OMS to store both the 3D model and the associated meta

information is that the graphics content and the semantics of the scenes are combined into a

consistent and cohesive ontological model. Users can connect objects in a 3D scene with classes









in a domain ontology. By doing this, the 3D scene becomes enriched with semantic information.

Associating domain ontologies with 3D shapes also generates a taxonomy which can be

navigated using a taxonomy browser such as the Web taxonomy Browser [14].Using an ontology

management system also opens up several other possibilities as discussed in chapter 5, such as

using reasoners to search the semantically augmented 3D greenhouse, positioning the greenhouse

on a geo-referenced 2D map and exporting the 3D model to other formats. Sharing models is

simplified since anyone can connect to the database remotely.

The text based search functionality is similar to the hyperlink based navigation with the

difference that the user viewpoint gets reset within the 3D model to the plant of interest. The user

can enter search terms and have their viewpoint reset within the 3D model to the region of

interest










CHAPTER 4
ONTOLOGY BASED 3D VIRTUAL GREENHOUSE

Domain Ontologies

A greenhouse ontology is shown below in figure 4.1.



LTength Width Ht at middle


Fhgas3DModel d Greenhouse
KW Height
NKW mat
Attached Structure Free Standing structure SKW mat

EKW mat
WKW mat
Lean To -- Quonset Type oi


Lean-tos are useful where space is limited to a width of approximately seven to twelve feet andmat
Full Span -f a L -s ae Arched Roof Type elect
SW mat --WW mat


Window Mounted -- Gable Roof Type



Figure 4-1 Greenhouse domain ontology. Grey shaded boxes represent classes. Arrows point
from a subclass to its parent. Blue shaded boxes are properties.

A lean-to greenhouse is a half greenhouse, split along the peak of the roof, or ridge line.

Lean-tos are useful where space is limited to a width of approximately seven to twelve feet and

they are the least expensive structures. The ridge of the lean-to is attached to a building using one

side and an existing doorway, if available. Lean-tos are placed close to available electricity,

water and heat. The disadvantages include some limitations on space, sunlight, ventilation, and

temperature control. An even-span is a full-size structure that has one gable end attached to

another building. It is a costlier option and it provides more space than a lean-to greenhouse.

Window mounted structures provide a convenient method for growing a few plants at relatively









low cost. The disadvantage with most of the attached greenhouses is that if not positioned

properly, exposure to the sun may be inhibited by the attached structure. There are also problems

with snow or water sliding off from the roof of the attached structure onto the greenhouse and

with walls of the attached structure retaining the sun's heat while the greenhouse cover loses heat

rapidly.

Freestanding greenhouse structures are separate structures. They can be set apart from

other buildings to get more sun and can be made as large or small as desired. The Quonset frame

style is a simple and efficient construction with the disadvantage that the sidewall height being

lower, storage space and headroom is restricted. As far as structural components are concerned,

the Quonset has no side walls compared to the arched and gable style greenhouses. All the three

freestanding structures may or may not have kneewalls.

The Greenhouse class forms the root of the taxonomic hierarchy. 3 attributes describe its

structure, namely, length, width and height at the center. Attribute restrictions on each attribute

specify that the minimum and maximum cardinality values to be 1 meaning that every

greenhouse instance should have these three attributes exactly once. The greenhouse class also

has a restriction on an attribute called "has3dmodel" .This Attribute restriction sets minimum

cardinality to 0 for attribute "has3dmodel" and it is used to link a greenhouse instance with the

Lyra object that contains the 3D rendering information.

Freestanding structures may or may not have knee walls. If present, there may be different

materials for the north, south, east and west knee walls. This is reflected by creating attributes,

one for each knee wall and adding Attribute restrictions to the Freestanding structures class to set

the minimum cardinality to be 0 and maximum to be 1 for each of these attributes. A domain

ontology for tomato diseases was created and stored in Lyra. [14] was used to create the









ontology shown in figure 4.2. A 3D object in the virtual greenhouse will be associated with a

class type in the domain ontology. By doing this, the 3D scene gets transformed into an

environment which knows about its 3D objects and can be queried using reasoners to infer facts.


Figure 4-2 Tomato Plant ontology. Light Gray denotes classes, dark grey denotes
instances, and light yellow denotes properties.

Database Design for 3D Objects

VRML Shape nodes are the only type of nodes for which physical entities are created in

Lyra. The translation, rotation and scale values are stripped off form the parent transform nodes

and stored as attributes with the shape node. Attribute vectors are created, one for each shape

node present in the VRML file. Each attribute vector records the translation, scale and rotation









values inherited from all parent transform nodes along with the coordinate and texture coordinate

values. Each attribute vector is named after the corresponding node's DEF name if the node has

one or a random name based on the nodes position in the hierarchy All the attribute vectors

created are then grouped together and added to the greenhouse instance as an attribute instance

of type "has3DModel".

User Interface

The user interface shown in Figure 4.3 is used to construct a greenhouse. The user can

specify properties such as the length, width, height and the temperature at which the greenhouse

must be maintained. By default, the greenhouse is made out of glass. The users can also click on

the "expand" button and change the type of structural materials for each part of the greenhouse as

shown in the Figure. In addition, the user has to specify a VRML file that can be used to

visualize the greenhouse in 3D. The user can save the configuration or load an already serialized

configuration from the File menu.

Methodology

For each greenhouse that the user creates, an instance is created in the database with the

user specified name. The instance refers to the one of the different class types specified in the

domain ontology as its parent depending on the greenhouse type that the user chooses. Attribute

instances are created to store the length, width ,height, temperature and other properties of the

greenhouse and they are added to the greenhouse instance. The VRML parser is run on the

VRML file to create attribute vectors that contain the rendering information .The attribute

vectors are linked to the greenhouse instance by adding it as an attribute instance of type

"has3DModel" to the greenhouse instances To view an already serialized greenhouse, the user

loads the greenhouse using its name. Lyra is searched for an instance with the user specified

name. The instance, if present, is obtained and its attribute instances are extracted. Using the data










present, the user interface is then populated. Using the meta information details extracted from

the database, a meta information map is created.




~Sa ViRTUAL. MODE tBIER
CG.MEEUOM o= lTKORNAYFON



i ypi OtrM nH. |Qu.l...
I CmSpZbe I I CtmnpIc
MATI*JAL RJORMATICd
N tVur Nmanrial IAWL- w SWau MFNtrtil IMEiA -I
E Wmam IFkrhril IGlaIs WJ 1M,1A Maer al Gldas
NHanee ilerll ILnraice lloInk ] th naeM material a i t, can m th a
E K atrmUHnd Crcielstock WImmU uMte ri n e1 ,k
IRT of Merl cl assi




Figure 4-3 Swing based UI for setting the properties of the greenhouse.

The map contains two types of key value pairs. The first type maps a DEF name to a user

defined name and the second type maps a user defined name to relevant meta information. This

map is kept synchronized with the database and any changes made to the map are immediately

updated in the database.

Mentor Mode

Clicking on the "mentor mode" option in the visualize menu brings up the 3D user

interface. The 3D model is rendered and displayed as shown in 4.4 depending on the greenhouse

type the user selects in the user interface.There are several ways users can navigate through this

model. Walking enables collision, so if the user collides on a wall or an object while moving he

is stopped. The "Fly" mode turns off collision detection. In addition, users can also pan through

the 3D model or tilt the model to adjust their visualization angle. Context clicking on the 3D









model brings up a menu. The users can search the greenhouse or inspect parts of the greenhouse

by clicking on the corresponding menu item.

Methodology

When the user requests an entry into the mentor mode, the 3D model corresponding to the

greenhouse the user loaded earlier is pulled up from the database. Using Xj3D, shape nodes are

constructed from each of the attribute vectors. A touch sensor node and a viewpoint node are

created for each and every shape node. A custom event handler is created and associated with the

touch sensor node. The event handler remembers the DEF name of the shape node and provides

a call back method that gets fired when a user touches this shape node. The call back method

creates a dialog that shows meta information about the shape node. A transform node is created,

one for each shape node and the translation, scale or rotation attributes are added as fields of the

transform node. The shape node and the associated touch sensor and viewpoint nodes are added

as children to this transform node. In addition, the sensor node gets added to a vector of sensor

nodes and the viewpoint node gets added to a hashtable which is keyed off the DEF name. The

transform nodes are then added to the scenegraph.

Inspect Mode

When in mentor mode, if the user wants to inspect objects in the 3D model, he/she enters

the "Inspect Mode". The user can do this by context clicking on the 3D model and selecting the

"Inspect Mode" option. By doing this, all the touch sensor nodes, one per shape node, that were

added during the time the scene was loaded,are enabled by getting each sensor node from the

vector and setting their "isEnabled" eventIn to true. The user can switch out of the inspect mode

any time by context clicking again and selecting the "Navigate Mode" option. When the user

goes back to the navigate mode, the touch sensors are disabled.
































Figure 4-4 The 3D greenhouse rendered in Xj3D.

In the inspect mode, when the user clicks on a component, a modal dialog is displayed.

Information in this dialog is divided by a tabbed pane. The user can enter the user defined name

for this component, add viewpoint metrics and textual comments describing this component in

the first pane as shown in Figure 4.5.Using the second pane, the user can add images relevant to

this component. Images that have already been added show up as thumbnail buttons before the

"Add a new image" button. Clicking on the thumbnail renders the full size image. Title for the

image can also be changed by clicking on the button next to it. New images can be added to the

existing ones on the fly by clicking on the "Add a new image" button. Clicking on this button

brings up a file dialog which can be used to locate the image. Once the image is selected, a

thumbnail of the image is created and added to the existing ones. Clicking on the save button

saves the new meta information. Figure 4.6 shows images of different types of heaters being

added.




































k i. 1l" '1- l' i.' .l l- I "" .. I ... I J. 11 = ,IIL -I i" Ir -

Figure 4-5 Textual information pane containing comments, user defined names and viewpoints.

Methodology

In the inspect mode, when the user clicks on a shape, the meta information map is searched

for the shape's DEF name to get the user defined name. The map is searched again based on the

user defined name obtained in the previous search. This search returns back all the meta details

associated with this shape which is then used to fill the meta information dialog with data.

When the user clicks on save in the meta information dialog, Lyra is searched for classes

or instances with the same SOID as the user defined name. Attribute instances are created and

user provided meta information like DEF name, comments and images are added as Data values

to the corresponding attribute instances. A new attribute called "refersTo" is used to link the

instance corresponding to user metadata with the class or instance in the domain ontology. The






























Figure 4-6 Image of an electric heater being added to the 3D object. The image shows the
thumbnail and the full sized image.

attribute, as its name implies adds semantic information by asserting that the 3D object that the

user clicked on maps to a particular class type or a particular instance in the domain ontology.

An attribute instance of type "refersTo" is created and it stores the OID of the class or instance in

the domain ontology as its data. All the attribute instances are then grouped together into an

attribute vector named after the user defined name and the attribute vector is added to the

greenhouse instance. Two entries, one with DEF name and user defined name as the key-value

pair and the other with the user defined name and OID of the instance created previously as the

key-value pair, are added to the meta information map.

For example, when a user clicks on the 3D object representing a tomato plant, the modal

dialog is displayed. The user then tags the object with the user defined name "Bacterial Wilt On

Tomatoes", types out a short textual description and adds some images. When the user clicks on

save, Lyra is searched for an object with the SOID "Bacterial Wilt On Tomatoes" which is an

Instance in our domain ontology. An attribute instance of type "refersTo" is created and the OID









of the "Bacterial Wilt On Tomatoes" instance is added as the data value to the attribute instance.

Attribute instances are created to store comments and images. All the attribute instances are then

grouped together into an attribute vector and updated back to the current database instance. Two

entries are made in the meta information map ,one being the key value pair

and the other being the pair <"Electric Heater",OID>.

By this mechanism, the 3D scene gets enriched with the semantic information present in

the domain ontology. It is possible to use a reasoner in the future, as discussed in the next

chapter, to make inferences or to process queries like "Show me all plants in the greenhouse that

are diseased".

The dialog also provides a way by which the user can group different 3D objects into a

single entity. For instance, the 3D model for a bench contains several shape nodes one for each

face of the bench. This can be verified by clicking on different faces of the bench and observing

that each face has a different DEF name. Using this tagging functionality, we can set the user

defined name to be the same for all parts of the bench thereby grouping the shape nodes under a

common name. Having a common name comes with the additional benefit that meta information

added for one shape node of the group shows up when any other member of the group is

inspected. Figure 4.7 illustrates this.

Searching

Clicking on the search menu item brings up an input dialog in which the user enters the

user defined name of the component being searched. Figure 4.8 depicts searching for corn plant.

Methodology

Searching for a component with its user defined name requires that the component be

tagged with that user defined name. In addition, the viewpoint details for the component












META INFORMATION EDITOR

fac oftehaer ci.-'J t- _. p _,'" DE n R domhildl34, =d eq



OX: I OZ: R:
G rrm nts A bild bhits c tpaner a e Primer.y








META WORWTIION EDTOR





















face reflecting the changes made to the meta information of the side face since they
are tagged with the same user name
OX.. oz R: Z




















must have been added as a part of meta details. When the user enters a user defined name in the

search dialog, the meta information map is searched for the name and the viewpoint details for
name <




must have been added as a part of meta details. When the user enters a user defined name in the





that node are extracted from the data. Then the set name corresponding to this associated name is

found from the map. The hashtable of viewpoint nodes which was created during the loading

stage is searched using this set name as the key and the node relevant to the searched object is









obtained. The viewpoint position in the viewpoint node is set based on the meta information and

its "setbind" EventIn is set to true. This results in the viewpoint of the shape node being bound

to the scenegraph thereby re-positioning the user orientation.

Energy Estimation

There are two basic equations used to calculate energy flow through a greenhouse. One is

for conduction, the transfer of heat through the structure of the greenhouse, and the second is for

convection, the transfer of heat by mixing warm greenhouse air with cooler outside air. When

calculating the heat loss for any greenhouse structure, both of these equations are taken into

account, at least to some degree. The particular equations used are from [13], and are outlined

below.

To calculate conduction over any greenhouse surface, four values are used; the

temperatures on either side of the surface (usually the temperature set point of the greenhouse

and the current temperature outside), the area of the surface, and a constant known as the U-

value of the surface material. The U-value represents a typical amount of heat energy that will

cross a certain area of the material, per degree difference between the inside and outside

temperatures.

The amount of heat per hour that is lost to that surface can be found using the equation

Q = u A AT Equation 4.1

Table 5.1 shows different uValues for different materials that the model supports. For

example, if a small greenhouse was set to 65F and the outside temperature was 45F, a 100 ft2

sheet of glass with a U-value of 1.13 BTU/ft2 -F-hour, then the amount of heat lost over the

sheet of glass would be: (65F 45F) 100 ft2 1.13 BTU/ft2 -F-hour = 2260 BTU/hour. To

calculate convection for any particular greenhouse, four values are used; the temperatures on the

inside and outside of the greenhouse (the difference being AT), the volume of air within the









greenhouse (V), and a correction factor (Cp) representing the relative leakiness of the greenhouse

structure. An estimate of heat lost can be obtained using the equation

Q = V Cp AT Equation 4.2


Figure 4-8 Search demonstration. A) The user labels the 3D plant object with the user name
"corn". B) Shows the user searching for the corn plant. C) Viewpoint is reset to point
to the corn plant

For example, if the temperature outside was 45F, the temperature inside was maintained at

65F, the volume of the greenhouse was 5000 ft3 and the greenhouse was assumed to have an









average leakinesss' (correction factor = 0.0216 BTU/ft3-oF-hour), the amount of heat lost per hour

due to convection would be (65F 450F) 5000 ft3 0.0216 BTU/ft3-oF-hour = 2160 BTU/hour.

Calculations assume average leakiness and always use the correction factor value of 0.0216

BTU/ft3-oF-hour. To find an estimate of the heat energy lost for the entire structure, conduction is

calculated for each wall, each portion of the knee wall, and for the roof, and convection is

calculated for the whole structure. Adding the results together gives us the estimate. The values

used in the calculation of area/volume of the greenhouse are the structural parameters that the

users specify through the interface like length, width and height. A synthetic database of external

temperatures, one value per hour over a period of one year, provides the external temperature

needed in the calculations. A synthetic database of expected solar radiation was used to get an

estimate of how much heat the greenhouse will gain from the sun. The amount of light energy

hitting per square foot of ground space was multiplied by the total footprint of the greenhouse to

get the estimated energy gain from the sun.

Using the above calculations the net amount of heat energy leaving or entering the

greenhouse is calculated. This is simply done by adding the total heat lost to conduction and

convection to the heat gained by solar radiation. If the result is negative, i.e., the greenhouse is

cooling off, the number of BTUs lost is assumed to be the number of BTUs added to the

greenhouse by its heaters, to keep it at its current temperature. If the number is positive, it is

assumed that the greenhouse is being cooled somehow and that no fuel is being burned to heat

the structure.

Comparison Mode

The user loads an existing greenhouse or specifies the structural properties for a new one.

Then when a comparison is run with another greenhouse, cost computation is performed for both

the greenhouses. A 3D model corresponding to the greenhouse type is pulled up. When the user









clicks on a component, energy lost form that particular component over each time frame is

tabulated for both the greenhouses.

Table 5-1 Uvalues

MATERIAL TYPE U-VALUE
GLASS 1.13
PLASTIC 1.15
DOUBLE LAYER PLASTIC 0.70
CONCRETE BLOCK 0.51
POURED CONCRETE 0.75
INSULATED CONCRETE 0.13

Simulation mode

Clicking on simulate in the visualization menu loads up a 3D model of the greenhouse.

The users can walk into the greenhouse and see the status indicator on the heater change at run

time based on several factors such as the external temperature, the internal temperature or the

solar radiation. The indicator goes green when the heater gets switched on and goes red when the

heater gets switched off. In addition the digital thermometer on top of the heater updates once

every few seconds showing the internal temperature in the greenhouse. It can be observed that

the heater gets switched on when the internal temperature in the greenhouse falls below the set

temperature that was specified as a part of the greenhouse design. The users can also click on any

part of greenhouse and get energy loss statistics corresponding to that part.

Methodology

The 3D object representing the light bulb on the heater is tagged with the name "Heater".

The digital text that shows the internal temperature in the scene is a 2D text node. The text node

is tagged with the user defined name "Thermometer".

During the simulation, at every time step, the internal temperature is calculated for the next

time step using the equations described in the earlier. The 2D text node with the name









"Thermometer" and the shape node with the name "Heater" are extracted from the scene. The 2D

text node is updated with the current internal temperature by calling the "settext" eventIn of the

text node. Then based on the heater's state, the Heater node's color is update by calling the

"setcolor" eventIn of the shape node. The external temperature data used is from a synthetic

dataset that provides external temperatures for every 15 minutes over a period of 24 hours. 2

seconds in the simulation is equivalent to 15 minutes in real world. Figure 4.9 shows the

simulation.


Figure 4-9 Simulation with the heater showing the internal temperature in the greenhouse.









CHAPTER 5
CONCLUSIONS AND FUTURE WORK

Walk through simulations play a useful role in the understanding of design. Previous

methodologies adopted in building a virtual greenhouse focused on either providing just a 3D

model or on computer based simulations that captured property changes in the greenhouse due to

various factors. This research focuses on meeting both the objectives by coupling the 3D model

with its information source. Lyra, an ontology management system, was used as the data store.

Xj3D was used as the VRML browser and toolkit. Using Lyra helped in organizing domain

specific information by using ontologies and these ontologies provided conceptual maps to

which the 3D models and other meta information sources were attached. Marrying 3D shapes

with domain specific ontologies transforms the static 3D scene which can be used for

visualization alone into a semantically rich 3D environment with which multiple users can

interact, even from a remote location, by learning from existing information sources or adding

more meta information. The 3D visualization also improves the search experience by allowing

the user to change viewpoint from the current location to the destination. This creates an

experience that is closer to the real world than navigating static web pages using hyperlinks.

A simulation of energy loss from the greenhouse was developed. The 3D visualization

enables the user to see the energy loss from each part by clicking on it. Comparison between

greenhouses that have the same structure or comparison of a greenhouse configuration with its

older version helps isolate areas of energy loss.

Ontology based search

Lyra being an ontology management system supports defining ontologies as we already

discussed. Using Lyra to store the meta information and the 3D models opens up several

possibilities related to searching the model. The VR environment can be considered as projecting









of objects described by the ontology into a 3D visualization environment. The ontology provides

a way of describing object properties and behaviors, describing object taxonomies, and, using

ontology reasoners, automatically classifying and cluster objects into categories. The ontology

acts as a dictionary that defines the meaning of concepts, and provides support for natural

language references to objects.

Ontology reasoners compare the structures of objects to determine how they are the same

or different. They can automatically classify new concepts and build categories by grouping

together similar objects. This capability leads, among other things, to query processing, as a

query can be classified in order to identify objects related to the query (e.g. find all instances that

satisfy the constraints of a query). In VR environments, such reasoning processes have many

applications. For instance, a user could pose a query such as "Show me all free standing

greenhouses in this database". The reasoner would infer that Quonset, arched and gabled style

greenhouses are all free standing structures. It would then get the instances of each class type and

display a list showing the names of each instance to the user. The user could then select the

greenhouse of his/her choice and open up its configuration. Another example would be a query

like "Show me a plant that is diseased". The reasoner sees that there is an instance which

"refersTo" the "powdery mildew on tomatoes" instance in the domain ontology. The reasoner

infers that powdery mildew is a type of tomato fungi which in turn is a type of tomato disease

and it hence classifies this instance as a diseased tomato variety. It gets the user specified name

for this diseased instance and loads it up from the database. It gets the viewpoint data and

repositions the camera to point to the 3D shape that represents the diseased tomato plant.

Natural language Support

Support for natural language queries provided by the ontologies enables objects to be

named or referenced in many different ways. Multilingual naming and facilities for managing









synonyms and homonyms enabled the same object to be named and referenced many different

ways. The Lyra OMS includes facilities for creating dictionaries and storing phrase patterns

needed to process more complex definite descriptions in natural language. An example of this

would be the way the user interface renders the comments and meta information to the user.

Based on the user requirements and locality, the UI could leverage the support provided by Lyra

to store the textual description in multiple languages.

Geo referencing the 3D model

Most of the virtual tours and virtual greenhouses on the Internet contain information about

where the real world peer is physically located. As illustrated in the literature survey, the Virtual

Field Day project uses a Flash Object to show the spatial relationship between different

greenhouses, ponds and fields. Support for georeferencing a greenhouse can be done relatively

easily.

ESRI, a company that specializes in GIS technologies uses Shape files (.shp) to represent

geospatially referenced maps [23]. A parser that works off the shape file has already been

implemented. The parser is similar to the VRML parser in its functionality. It works on the shape

file and stores the geographic information in Lyra. The data can then be used to generate a flat

2D map of the locality. The 3D model can be transformed, scaled and placed at its coordinates.

Thus, the user can actually see the georeferenced map and the 3D model placed on the map

giving a better viewpoint of the landscape.

An example of this is the ISAS project[18]. A geospatially referenced campus map was

imported into the OMS by parsing ESRI Shape Files and a detailed 3D model for one of the

buildings at the University Of Florida (Rinker Hall) was positioned at its coordinates. A similar

technique could be used for positioning greenhouses in real environments, such as an existing

nursery operation.
























Figure 6.1 Images showing geo-referenced Rinker Hall. A) Front view B) Top View shows the
entire map

Exporting the 3D model to other formats

Storing the 3D model in Lyra makes it possible to export it into a variety of other formats

since Lyra provides a clear separation between content and presentation. For instance, the 3D

coordinates in the form of attribute vectors in Lyra can be exported to a VRML file or an X3D

file. The exporter would be a software component that operates within Lyra. The target file

format is created by the exporter module and this makes it possible to plug-in different exporters

to Lyra to output the 3D content in different file formats.

Thus, the virtual greenhouse system developed can be expanded to accommodate several

such future extensions very easily.









LIST OF REFERENCES


[1] Stacy, J. & Shubat, D. & Anna, T. (2003).Greenhouse Virtual Tour.
University Of Minnesota. http://www.d.umn.edu/biology/greenhouse/VirtualTour/
VirtualTourOfUMDGreenhouse.htm

[2] IFAS. (2006). Virtual Field Day. University Of Florida. http://vfd.ifas.ufl.edu/

[3] Tignor, B. et al. (2007). Greenhouse Simulator. University of Arizona.
http://ag.arizona.edu/ceac/wge/simulator/greenhouse_content.html

[4] Tignor, B. et al. (2007). Worldwide Greenhouse Education. University of Arizona,
University of Florida, University of Vermont, Ohio State.
http://www.uvm.edu/wge/index.html

[5] Frantz, J. & Buckingham, L.A. & Locke, J.C. & Krause, C.R. (2006). Virtual grower.
http://www.ars.usda.gov/services/software/download.htm? softwareid=108

[6] T. Honjo & E. Lim. (2002). Visualization of Environment by VRML. in Proceedings of
the World Congress of Computers in Agriculture and Natural Resources.

[7] Aono, M. & Kunii, T. (1984). Botanical Tree Image Generation. IEEE Computer
Graphics & Applications.

[8] VRML. (1995).Virtual Reality Modeling Language. World Wide Web Consortium
http://www.w3.org/MarkUp/VRML/

[9] ECMA. (1999). ECMA Scripting Language Specification. ECMA-International.
http://www.ecma-international.org/publications/standards/Ecma-262.htm

[10] Marrin, C. (1997). External Authoring Interface Specification. SGI.
http://graphcomp.com/info/specs/eai.html

[11] Xj3D (2006). Xj3D VRML Toolkit Specification. http://www.xj3d.org/

[12] Beck, H.W. (2005). Lyra Ontology Management System. University of Florida.
http://orb.at.ufl.edu/Obj ectEditor

[13] Aldrich & Bartok. (1994). Greenhouse Engineering, NRAES publication 33

[14] Beck, H.W. (2005). Web Taxonomy Browser. University of Florida.
http://orb.ifas.ufl.edu/index.html

[15] R. A. Bucklin. (2007). Physical Greenhouse Design Considerations. IFAS Extension.
University of Florida. http://edis.ifas.ufl.edu/CV254.









[16] Ross, D.S. (2004). Planning and Building a Greenhouse. University of Maryland
Cooperative Extension Service Fact Sheet 645.
http://www.wvu.edu/-agexten/hortcult/greenhou/building.htm

[17] R. A. Bucklin. (2002). Florida Greenhouse Design. IFAS Extension. University Of
Florida. http://edis.ifas.ufl.edu/AE016.

[18] Oliverio,J. & Masakowski, Y.R. & Beck, H.W. & Appuswamy, R. (2007). ISAS: a
human-centric digital media interface to empower real-time decision making across
distributed systems. In the proceedings of 12th international conference on 3D web
technology.

[19] Bartok, J.W. (2005). University of Connecticut
http://www.umass.edu/umext/floriculture/factsheets/greenhousemanagement
/jb building_gh.htm

[20] SL. (2007). Second Life. Linden Research. http://secondlife.com/

[21] Smajstrla, A.G. & Zazueta, F.S. (2002). Evaporation Loss During Sprinkler Irrigation.
IFAS Extension. University of Florida. http://edis.ifas.ufl.edu/AE048

[22] Java Swing (2007). Sun's Java VM. http://java.sun.com/

[23] SHP File Format (1998). ESRI SHP File Format.
www.esri.com/library/whitepapers/pdfs/shapefile.pdf









BIOGRAPHICAL SKETCH

Raja Appuswamy received his Master of Science in Computer Engineering and

Agricultural and Biological Engineering from University of Florida in Fall 2007. He worked

with Dr.Howard Beck and professor James Oliverio on ISAS and published a paper titled "ISAS:

A Human-Centric Digital Media Interface to Empower Real-Time Decision-Making Across

Distributed Systems" in Proceedings of the twelfth international conference on 3D web

technology.He pursed research on Virtual Greenhouses for his Masters thesis under Dr.Howard

Beck and Dr.Ray Bucklin. Raja received his Bachelor of Engineering degree in Computer

Science from Anna University, India in 2005





PAGE 1

DESIGN AND DEVELOPMENT OF ONTOLO GY-BASED 3D VIRTUAL GREENHOUSES By RAJA APPUSWAMY A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2007 1

PAGE 2

2007 Raja Appuswamy 2

PAGE 3

To my mom dad and friends 3

PAGE 4

ACKNOWLEDGMENTS I thank Dr.Howard Beck, my advisor and chai r, for all the support he has offered me during the course of my studies as a graduate student. His style and approach to problems profoundly influenced me and helped me hone my technical knowledge. This research would have been impossible wi thout the support offered by Dr.Ray Bucklin. I would like to thank him profusely for guiding me and providing me with the right resources at the right time. I am indebted to Professor James Oliverio as my experience working with him on ISAS greatly influenced the design of the 3D virtual greenhouse. I would like to thank my parents for I would have never made it this far without their blessings and last but not the least, I would like to tha nk all my friends whose constant encouragement and support made everything possible. 4

PAGE 5

TABLE OF CONTENTS page ACKNOWLEDGMENTS ...............................................................................................................4LIST OF TABLES ...........................................................................................................................7LIST OF FIGURES .........................................................................................................................8ABSTRACT .....................................................................................................................................9 CHAPTER 1 INTRODUCTION ................................................................................................................ ..112 LITERATURE REVIEW .......................................................................................................132.1 UMD, UWMadison Greenhouse Virtual Tour ............................................................132.2 University Of Florida Virtual Field Day .......................................................................142.3 The Greenhouse Simulator by University of Vermont ..................................................142.4 Virtual Grower ...............................................................................................................162.5 Visualization of Environment by VRML ......................................................................172.6 Evaluation ............................................................................................................. .........183 TOOLS AND TECHNIQUES ................................................................................................20 3.1 Java Swing .....................................................................................................................203.2 VRML ............................................................................................................................213.4 Xj3D ..............................................................................................................................243.5 Lyra Ontology Management System .............................................................................253.6 Addressing the Limitations ............................................................................................304 ONTOLOGY BASED 3D VIRTUAL GREENHOUSE ........................................................334.1 Domain Ontologies ...................................................................................................... ..334.2 Database Design for 3D Objects ....................................................................................354.3 User Interface .................................................................................................................36 4.3.1 Methodology ......................................................................................................364.4 Mentor Mode .................................................................................................................37 4.4.1 Methodology ......................................................................................................384.5 Inspect Mode .................................................................................................................38 4.5.1 Methodology ......................................................................................................404.6 Searching .......................................................................................................................42 4.6.1 Methodology ......................................................................................................424.7 Energy Estimation ...................................................................................................... ...444.8 Comparison Mode .........................................................................................................464.9 Simulation mode ........................................................................................................ ....47 5

PAGE 6

6 4.10 Methodology ................................................................................................................475 CONCLUSIONS AND FUTURE WORK .............................................................................495.1 Ontology based search ...................................................................................................495.2 Natural language Support ..............................................................................................505.3 Geo referencing the 3D model .......................................................................................515.4 Exporting the 3D model to other formats ......................................................................52LIST OF REFERENCES ...............................................................................................................53BIOGRAPHICAL SKETCH .........................................................................................................55

PAGE 7

LIST OF TABLES Table page 5-1 Uvalues ..............................................................................................................................47 7

PAGE 8

LIST OF FIGURES Figure page 2-1 Web page showing the 360 de gree Quicktime virtual tour. ...............................................15 2-2 Greenhouse simulation ..................................................................................................... ..16 2-3 Images of Virtual Grower .................................................................................................. 17 2-4 VRML Greenhouse developed at Chiba University ..........................................................19 3-1 Domain ontology for Computers .......................................................................................30 4-1 Greenhouse domain ontology.. ..........................................................................................33 4-2 Tomato Plant ontology.. ................................................................................................... ..35 4-3 Swing based UI for setting the properties of the greenhouse. ...........................................37 4-4 The 3D greenhouse rendered in Xj3D. ..............................................................................39 4-5 Textual Information Pane containing comments, user de fined names and viewpoints. ....40 4-6 Image of an electric heater being added to the 3D object. .................................................41 4-7 Aggregating shape nodes under a common user defined name. ........................................43 4-8 Search demonstration. ..................................................................................................... ...45 4-9 Simulation with the heater showing the internal temperature in the greenhouse. .............48 6-1 Images showing geo-referenced Rinker Hall. ....................................................................52 8

PAGE 9

Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master Of Science DESIGN AND DEVELOPMENT OF ONTOLO GY BASED 3D VIRTUAL GREENHOUSES By Raja Appuswamy December 2007 Chair: Dr. Howard Beck Major: Agricultural and Biological Engineering With the growing popularity and widespread appreciation for the design and construction of greenhouses, several attempts have been made over the past few years in developing web based virtual greenhouses. The approach taken towards the design of such greenhouse can be classified into two types. The first type focuses on delivering knowledge to the users. Visualization is either not a part of such a syst em or is minimal. The second type is centered on developing a visualization of the greenhouse so that the users can get a feel of the real world peer. Such a visualization system stands alone as a separate entity and is not linked with any meta information. This research focuses on developing a 3D virt ual reality based virt ual greenhouse. The 3D model is coupled with information sources such as text based descriptions and relevant images by using a single data store (Lyra OMS) to store both the 3D model and meta information. Using Lyra made it possible to link objects in the 3D scene with domain ontologies and thus build a rich 3D virtual environment in which users can create new greenhouses, save and load previously saved greenhouse instances from the database, walk through the 3D greenhouse model, inspect 3D objects and view all relevant information. They are able to tag objects with 9

PAGE 10

10 names and group multiple 3D objects with the same name. A search functionality that redirects the user to his/her point of interest within the 3D sp ace has also been implemented. An energy loss estimation engine operates on the user specified properties of the greenhouse and produces statistics of energy loss per structural component. Users can compare the energy loss characteristics of multiple greenh ouse configurations and refine their design. An educational simulation that shows the status of an electric heater with the internal temperature has been implemented to illustrate the capability of extending this system to support simulations.

PAGE 11

CHAPTER 1 INTRODUCTION Planning, designing and constructin g a greenhouse has become an activity of interest to a wide range of people from hobbyists to horticultur al experts. To encourage people to develop greenhouses and to educate people worldwide about developing such structures, the concept of developing a virtual greenhouse has become ve ry popular. These virtual greenhouses are web based electronic representations of their real worl d peers. The purpose of such greenhouses is to provide a repository of information which users can utilize to make quality decisions about constructing their own greenhouse or selecting which plants to grow in what types of greenhouses. Some virtual greenhous es link to diverse multimedia information sources like images, educational videos and Flash animations. The user navigates through the model and follows the links to his/her area of interest. With the advent of 3D on the Internet, one of the ways to model a real world greenhouse is to construct a model in 3D. Several virtual reality (VR) environments exis t that import such 3D models and render them. Using these VR environm ents, the users are able to walk through the virtual greenhouse and look at plan ts or other components at a various levels of detail. Traditionally 3D virtual environments util ize files (VRML, X3D) that describe the geometry of 3D objects but contain very little additional knowledge of the objects in the VR world. There are many reasons to enhance VR envir onments with better database support. In a VR world, we need to know not only what an object looks like, but also wh at the object is, what its properties and characteristics ar e, how it behaves, and how it rela tes to other objects. We also need to be able to interact with and refe r to objects by pointing or by voice commands. In our approach we utilize an ontology management system, a database management system based on ontologies, to provide semantic descriptions of objects in the VR environment. 11

PAGE 12

12 In fact the VR environment can be considered a projecting of objects described by the ontology into a 3D visualization environment. The ontolog y provides a way of describing object properties and behaviors, describing obj ect taxonomies, and, using ontol ogy reasoners, automatically classifying and clustering objects into categories. The ontology acts as a dictionary that defines the meaning of concepts, and provides support fo r natural language references to objects. This thesis focuses on studying th e practicability of a virtual reality based 3D environment which is augmented with meta information. The purpose of this research is to build a unified VR environment which offers its users the capability to walk through the 3D model and also inspect, add, modify and search for meta information on the fly. The meta information provided by the user is then used to relate objects in the 3D scene with dom ain ontologies. Tagging 3D shapes with ontologies builds up a taxonomy that can be navigated by the users to find relevant information. In addition, coupling semantic cont ent with the 3D shapes also makes future extensions possible, such as using a reasoner to replace a text based search with semantic search.

PAGE 13

CHAPTER 2 LITERATURE REVIEW University of Minnesota Duluth (UMD), Un iversity Of Wisconsin Madison Greenhouse Virtual Tour UMD virtual tour [1] was developed to expose the users to the different groups of plants in the UMD greenhouse and to help the users unde rstand the way the plants and benches are organized. The tour was simulated by a collect ion of web pages, each providing in-depth information about the greenhouses. The web site contains an index page that provided the visitors with general information and contai ns links to the two UMD greenhouses, the Upper Greenhouse and the Lower Greenhouse. Users could na vigate to the greenhous e of their interest by following the link. Each virtual greenhouse is modeled as a web page using HTML. The web page contains an image that shows the floor plan of the greenhouse. The fl oor plan describes the location of plants, benches and other fixtures within the green house and is image mapped to internal links within the web page Users can click on individual pa rts of the floor plan and jump to the content of their interest, or they could simply scroll down and view the entire content. Information about each plant, bench and fixture comes in two forms: a paragraph of textual information that describes the relevant informa tion and photographs pertinent to that component. The design of the UWMadison greenhouse is si milar to that of UMD. Use of hyperlinks to model the relationship between rooms in a greenhouse was the adopted technique. The index page contained hyperlinks to all rooms within the greenhouse. Each room was modeled as a web page that contained its own set of hyperlinks that linked to individua l plants within the greenhouse. Each individual plan t had a web page that provided detailed information about the plant and relevant pictures. 13

PAGE 14

14 University Of Florida Virtual Field Day The Virtual Field Day project [2] undertaken by the University Of Florida aims at providing the users access to a repository of information about the research greenhouses at Gainesville and Suwannee. The Virtual Field Day pr oject provides its users with an "Interactive Map" which is a Flash object embedded in a web page. The users can view the greenhouses in the map. The map assists the users in naviga ting around the locations in the Suwannee or Gainesville Virtual Field Day. Selecting a loca tion takes them to that location's multimedia resources. In addition to the text based and image based information, the site's multimedia resources include instructional videos, publications and research literature pertinent to the technologies used in the greenhouse. In addi tion, each greenhouse has a Quicktime VR tour which enables the users to interact with the 360 degree panoramas. The tour plays in Quicktime player and has tags on specific areas of interest. Navigation th rough the tour is limited to a stationary 360 degree view and zooms in/out. Clic king on tags takes the user to a web page that contains a listing of related videos, publications and other instructiona l material. Separate 360 degree views show the users how the greenhouse looks in the pre-plant season and during the main season. In both the Virtual Field Day project [2 ] and the greenhouses by UMD [1] and UW, photographic images serve as the primary m eans by which the users can visualize the greenhouse. All the information is static and th e users cannot simulate changes in the greenhouse environment based on properties like climate or st ructure. User interactivity is limited to the traditional web based clicking and hyper linking. The Greenhouse Simulator by University of Vermont The Greenhouse simulator [3] is a part of a world wide greenhouse educ ation initiative [4] undertaken by The University of Vermont, The Univ ersity of Florida, Un iversity of Arizona

PAGE 15

Figure 2-1 Web page showing the 360 degree Quicktime virtual tour on the right and instructional videos related to seeding, transplants etc on the left Virtual Field Day Project at University Of Florida [2]. and Ohio State. The outcomes of this project in clude a greenhouse simulato r, a digital repository and educational materials. The di gital repository contai ns a searchable collection of images and videos. The simulator is a Flash object that al lows users to model the greenhouse environment by making structural, environmental, geographic, a nd seasonal choices. The simulator incorporates user-selected information from its database of greenhouse designs, operation, and geographic climate conditions, and graphically displays dynamic changes in greenhouse environments, including moist air properties. The animation al lows learners to simulate changes in the greenhouse-plant environment based on climate, st ructure, and environmental control choices. The interactive greenhouse environment simulator was developed by integrating mathematical models and an animation inte rface, which was created using Flash MX Pro 2004.The greenhouse mathematical model, which is based on the energy and mass balance of the greenhouse system, utilizes a series of differen tial equations. The solution provides the dynamic response of the greenhouse climate conditions to the outside climate conditions and for a particular greenhouse design. The design incorporates user-selected inputs for climate, structure, 15

PAGE 16

glazing, and environmental control systems. Each simulation demonstrates the response of a greenhouse system design over a 28-hour period. The simulator is a stand-alone piece of software and is not linked to any other information sources. The user interface limits the interaction to the user selection of various parameters. Further information about components inside the greenhouse cannot be obtained from the greenhouse model in the simulation. Figure 2-2 Greenhouse simulation shown plotting valu es of solar radiation, air temperature, humidity etc over a period of 28 hours based on user specified inputs. [3] Virtual Grower Virtual Grower [5] is a PC based stand-al one program developed by the ARS Greenhouse Production Research Group at Toledo, Ohio. It is a decision support tool for greenhouse growers that enable the users to estimate the energy re quirements of a greenhouse. To estimate energy requirements and costs using Virtual Grower, users input the dime nsions of their greenhouse and its construction materials, such as poured-concret e floor, glass sides and r oof, or concrete-block walls. They also choose design featur es, such as roof shape and orie ntation to the sun. A historic database gives a year's worth of typical weathe r for the city nearest to the greenhouse location including factors such as temperature, sunli ght and cloud cover for each hour of the day. 16

PAGE 17

Growers choose a heating schedule and set the temp eratures they want to maintain during day and night, or for each hour. The program then calculates per-square-foot heating costs by the month or year. The users can sa ve their heating schedule, gr eenhouse configuration or export their results to other formats. A B Figure 2-3 Images of Virtual Grow er[5] A) User setting various parameters of the greenhouse B) Virtual Grower reporting the estimated heating costs The visualization of the gree nhouse constructed by the users is limited to static line diagrams that depict the struct ure of the greenhouse as shown above. The primary purpose of this software is cost estimation and hence visualizatio n is not a component. User interaction is also limited to selecting the various parameters. Visualization of Environment by VRML Honjo. T and Lim. E [6] designed a visualiz ation system for greenhouse that enables virtual experience in a planned environment was developed using VRML. Their research focuses on walk through simulations as a tool for understanding the design of landscapes and greenhouses. 3D plants and objects with a high po lygon count generally take a very long time to render and require expensive, hi gh power hardware for a rich walk through experience. To avoid 17

PAGE 18

this problem, a plant was expressed by using two textured planes instead of thousands of polygons. Computer generated images of plants at several growth stages are made by using AMAP (Atelier de Modelisation de Architecture de Plants) [7], which is the system developed by CIRAD ( Center Internationale Recherche Agri cultural Development). The plant images then serve as texture for the two planes thereby rendering the plants with minimal overhead. The greenhouse structure itself was ma de out of simple VRML constructs. The VRML virtual greenhouse enabled walk through. The 3D model was also posted on the web so that users could use a web browser to view the VRML world. Having the user walk through the 3D model provides a better unders tanding of various parts of the model since its equi valent to his/her walking through the real world greenhouse. This system focuses only on the visualization aspect. It does not relate 3D models with relevant information or other multimedia sources like images or educational videos. The 3D scene is static since it comes off a VRML file. The user cannot make changes to the 3D scene at runtime or add information to certain components. For in stance a user who has an image of a corn plant will not be able to add it to the collection of existing images. Evaluation Research literature indicates the need for a sy stem that merges the pros of both 3D based virtual greenhouses and other HTML based models While the 3D greenhouse provided a better user interaction and navigation capability, it was not associated with multimedia information sources. Compared to this, the traditional web based greenhouses had images, videos and text based information sources. The disadvantage with the HTML based virtual greenhouses was that visualization of the greenhouse was limited to photographs and walking through the greenhouse to foster a better understanding of the real worl d greenhouse was not possibl e with plain images. 18

PAGE 19

19 Figure 2-4 VRML Greenhouse developed at Chiba University [6]

PAGE 20

CHAPTER 3 TOOLS AND TECHNIQUES The following design goals were formulated: Build a 3D model of a greenhouse through which the users can walk. Provide a mechanism by which plants and ot her components in the 3D model can be annotated with meta-information and also insp ected to view existing information sources. Store the model and meta-information in an ontology management system that can be accessed remotely. Implement a search functionality so that the us ers can search and be redirected to see the objects of their choice. Develop an educational simulation involving thermal characteristic s of the greenhouse. The simulation will provide visual cues that depict switching a heater on and off based on heat loss from the greenhouse. This a dds dynamic behavior to the 3D greenhouse environment. To meet all the these goals, the follow ing tools and techniques were employed. Java Swing Java Swing [22] is part of the Java Foundati on Classes, a group of classes that encompass features for building GUIs and adding rich graphi cs functionality and interactivity to graphics applications. The other components that form a part of JFC include PLAF (Pluggable Look And Feel) Accessibility API to enable assistiv e technologies like screen readers Java 2D API to help developers enrich their applications with 2D graphics and text, Internationalization to enable the development of programs that can interact with the users in their own local language and cultural conventions Drag and Drop. Prior to Swing, Abstract Window Toolkit (AWT) was used to construct user interfaces. AWT had several problems associated with it. It was not sufficient for complex user interfaces, it was not portable, it heavily relied on the runtime platform to rende r many of its user interface 20

PAGE 21

21 components and had a poor event model. Swing was developed on top of AWT to solve many of the problems associated with it. Swing provides pluggable look and feel. All components in Swing are lightweight in the sens e that they do not depend on na tive runtime environment except for a few top level containers. Several new comp onents such as sliders and progress bars are added to the list of existing AWT components. This research uses Java Swing to build the User Interface. VRML VRML (the Virtual Reality Modeling Language) [8] is an international standard for describing 3-D shapes and scenery. VRML files describe a 3D world. Visualizing the 3D world can be done by viewing the VRML file in a VRML browser. Since many of the VRML browsers are available as web browser plug-ins, 3D info rmation can be easily publ ished on the web using VRML. The VRML file contains a node hierarchy wh ich reflects the way in which individual shapes in the 3D scene are grouped together to form composite objects. Grouping nodes in the VRML files serve to group together their children. Transform nodes, in addition to grouping a set of nodes also add a translati on, rotation or scale to all of its child nodes. Shape nodes contain information about the appearance and geometry of each shape in the 3D scene and are grouped within grouping nodes. The appearance of the sh ape node is usually provided by an Appearance node that controls several propert ies like coloring aspects, text ures, transparency and opacity. The geometry of a shape node can be described by making it one of the predefined types like a BOX or by using an IndexedFaceSet node. An Inde xedFaceSet node lists the coordinates of the points that form the shape and groups up points th at form an individual face of the shape, one group per face, thereby building up the structure. Nodes in a VRML file are named by using the DEF construct. Prefixing a node declaration with a user defined name a ssigns the user defined

PAGE 22

name to the node and makes it possible to refer to that node anywhere in the file using that name. USE construct makes use of such DEFed nodes to reuse existing nod e declarations. For example, in the following node hierarchy given below, an appearance node and a geometry node get DEFed with user defined names and they are reused further down the hierarchy to control the appearance and geometry of another shape. Shape { appearance DEF myAppearance Appearance { } geometry DEF myGeo geometry { } } Transform { translation 0 0 0 Shape{ appearance USE myAppearance geometry USE myGeometry } } The 3D scene built by the VRML nodes is st atic. To add animations or to make it interactive based on the user input, the VRML nodes must be connected to each other and messages must be passed between them. For this purpose, most of the nodes have special fields called eventIns and eventOuts. EventIn fields are like receivers, whic h receive messages called 22

PAGE 23

events from the outside and take them in to be processed. EventOut fields are transmitters, which send events from the node to the outside. Node s are connected to each other by attaching these fields using ROUTEs. These are like pi pes that channel events from an eventOut into an eventIn Many eventIns can be connected to one eventOut to make one event cause many things to happen. This is called fan-out Fan-in is also allowed, where two or more eventOut s feed into one eventIn VRML also contains some special nodes. Thes e nodes act like sensors in that they sense user interaction with the scene and react to it by throwing events. One such node that provides interactivity in the 3D greenhous e is the Touch Sensor node. This node detects mouse interaction with its sibling geometry. Just like other nodes, TouchSensor also has some eventIn and eventOut fields. isOver, isActive and touchTime are the eventOuts in TouchSensor. When the mouse is moved over the geometry, an isOver TR UE event is generated, and when it moves off, an isOver FALSE is sent. When any of the mouse buttons is clicked while isOver is TRUE, two events are generated, namely an isActive TRUE event and an the touchTime ( where currentTime indicates the time when the user clicked on the shape ) event is generated. To illustrate the process of creating animations in a VRML scene, consider a scene that has a touch sensor node and a S ound node. The Sound node is used in VRML for generating 3D sounds. When the touchTime eventO ut of the TouchSensor is connected to the beginTime eventOut of the Sound node using a ROUTE, the sound file specified in the sound node is played when the user clicks on the shape that the TouchSensor monitors. VRML files can be hand coded or generated from third party applications. The 3D models for the greenhouse are fairly complex. Hence, a third party software modeling tool SketchUp, was used to build the 3D model for each of 6 di fferent types of greenhouses. The models were 23

PAGE 24

then exported to generate VRML files. Xj3D VRML as such provides a language by which modelers can describe a 3D scene. VRML by itself cannot display a 3D scene without the soft ware to render the 3D scene. To render the environment, a VRML browser is required. Xj3D [11] is a project of the Web3D Consortium focused on creating a toolkit for VRML97. X3D c ontent is written completely in Java and provides a VRML browser that can be used to rend er a 3D scene. We can use Xj3D to render the VRML scene from a VRML file or use its API to build up the scenegraph directly by constructing nodes one at a time and adding them to the scene. Xj3D also has a parser implementation that can be used to parse a VRML file and generate an in memory representation of the VRML node hierarchy. Interactivity and animations can be added to the VRML file by two mechanisms. The first one is scripting within VRML. Scripting language s like ECMA Script [10], JavaScript can be placed inside special type of nodes called Script Nodes. These scripts process on internal events and change the nodes in the scenegraph. The ot her way is to use EAI (External Authoring Interface) [11].The EA I enables control of th e contents of a VRML browser window embedded in a web page from a Java (tm) applet or such an external source on the same page. The difference between EAI and intern al Script node is that EAI o ffers a generalized method to access nodes and events in the VRML browser's scene graph by connecting to a VRML browser plug-in through a browser plug-in architecture like Live Connect or ActiveX/COM. Internal Script Nodes have node access through definition within a VRML Script Node within a VRML scene graph. The EAI supports events and fields on a dynamic basis, while the Script Nodes definitions consist of predefined fields and events. Xj3D merges the EAI and scripting paradigm 24

PAGE 25

and provides a single unified in terface called SAI (Scene Acce ss Interface). It provides a compatible way to run and modify X3D and VRML scenes for many X3D implementations. The virtual greenhouse uses the Xj3D API for two purposes. Xj3D Parser API is used to parse the VRML file. The parser operates based on an eventing paradigm. The parser expects the caller to register a content handler with it. The content handler cont ains a list of functions that are called back by the parser at critical points such as document entry/exit, node entry/exit and field entry/exit. For instance, when the parser h its the beginning of a node it calls the function startNode on the content handler. The content handler contains a set of stacks that it uses to pass information between calls to the functions. For in stance, when the start Node function is called, the content handler creates a new hash map keyed on field names to store the field values for this node and pushes it onto the top of the stack. When fi eld data is encountered by the parser, it calls the fieldValue method which crea tes a new entry in the hash ma p. When a node definition ends, the parser calls the endNode method which remove s the hashmap from the top of the stack. Thus, using the parser API and a custom content hand ler implementation, an in memory representation of the VRML file can be generated. The second place where Xj3D API is used is to provide an environment for 3D visualization. Xj3D provides a VRML browser implementation that is used to host the 3D greenhouse. VRML nodes are constructed based on th e shape information present in the database and the nodes are added to the scene graph whic h renders itself once constructed completely. Thus, using SAI, the applicati on interfaces with the database and updates the scene at run time during the simulation. Lyra Ontology Management System In philosophy, ontology is a theory about the nature of existenc e, of what types of things exists; ontology as a discipline studies such theories. The most typical kind of ontology has 25

PAGE 26

taxonomy and a set of inference rules. The taxonomy defines classes of objects and relations among them. The most basic concepts in a domain form the roots of various taxonomic trees as classes. A concept more specific than the gene ral high level concept is considered to be a subclass. The generic concept becomes the superc lass of the sub class. Members of a class are called instances. Instances can be thought of as corresponding to individuals in the real world. Properties assert general facts about the members of cl asses and specific facts about individuals. Let us consider the plant ontology given below. Kingdom: Plantae Division: Magnoliophyta Class: Liliopsida Order: Cyperales Family: Poaceae Genus: Zea Species: Zea mays Kingdom, Division, Class, Order, Family and Genus are all classes representing concepts in this ontology forming a taxonomic hierarc hy for classification of plants. Plantae, Magniliophyta etc are instances of the correspondi ng classes. The kingdom Plantae indicates that corn is a plant and probably photosynthesizes using chlorophyll. The division Magnoliophyta indicates that it is a flowering plant. The class Liliopsida indica tes that corn is a monocotthat is, it has one cotyledon (the first l eaf formed on a seedling) and leaves with parallel veins, and its flower parts occur in multiples of three (three or six petals, three or six stamens, and so on). The order Cyperales indicates that th e seeds store starch and the flow ers lack petals and sepals. The family Poaceae indicates that it has a unique type of inflorescence, round stems, and a distinctive internal anatomy and so on, dow n to the species level. 26

PAGE 27

An ontology management system (OMS) is a da tabase management system that utilizes formal ontology languages as the basis for mode ling and manipulating data. The OMS provides facilities typically available in database manage ment systems including maintenance of database integrity, physical storage management, secur ity, transactions, bac kup/recovery, and data manipulation. Using the OMSs da ta definition language, onotologi es can be constructed, and using traditional data storage management techni ques, the ontology can be stored and retrieved. Physical storage management is particularly important because the VR environment requires management of very large numbers of objects. Dynamic scene generation requires that just the right objects to be loaded into memory at the appropriate time to expand a scene or increase level of detail in object presentation. This research uses Lyra [12], an Ontology Management System as a core database management facility. Lyra stores data in the form of objects. An object in Lyra is variable length block of bytes Each object can be identified using either an integer Object Identifier (OID) or a string Object Identifier (SOID). Lyra creates hash based indices on both OIDs and SOIDs to enable rapid lookup of objects. Lyras data definition language represents all generi c concepts as cla sses and individual members as instances. Attribute is the term used to represent a property in Lyra. Attributes are relationship between objects. They have a domain and a range and they are directional, meaning that they connect the source objects in the domain to the target objects in the range. For instance, an attribute made from wood could be used to connect a door in a gr eenhouse with a type of wood. Thus the domain would be class door and the range would be cla ss wood. Attributes can be used in either classes or in instances. When used with classes, attributes are used to restrict the set of values that the class instances can have fo r that attribute. Such a restriction imposed on the attribute values is called an attribute restriction. Cardinality is an example of an attribute 27

PAGE 28

restriction. Assigning a minimum an d maximum cardinality restricti on to an attribute enforces that the attribute appears at le ast minimum times and no more than maximum times in the class instances. For example, when an attribute restrict ion is created for the door class that specifies that the cardinality of the attri bute made from wood to be one, all instances of the door class must definitely specify exactly one va lue for the attribute made from wood. When an attribute is used in the context of an instance, it is called an attribute instance. attribute instance describe prope rties of an instance. For inst ance, the attribute made from wood could be used in an instance door1 to sp ecify that door1 is made from pine. Thus the attribute instance of type made from wood relates door1 with the wood instance pine. Lyra provides two ways for storing attr ibute instances in the database, namely, storing each attribute instance as an object or grouping together a set of attribute instances and embedding the group of values within an object rather than making individua l objects. Attribute vector is the term used in Lyra to represent the grouping of attribute instances. Each at tribute vector has a name and maintains a hashtable that stores all the attribute instances as a set. Lyra supports several data types from simp le ones like integers and strings to more complex ones like images and videos. The values that attributes take in an instance are stored in the Lyra as Data objects. A Data obj ect is actually a generic object type that is used in Lyra to store all types of data in binary format. Values are coerced and type casted to their corresponding data types as and when necessary. For example, consider the problem of representi ng a 3D point in Lyra. The coordinate of a point in 3D space can be represented using 3 valu es: x, y and z. Each of the 3 values is of floating point type. To model this in the form of Lyra objects, three attri butes representing the x, y and z properties of a point are created. Let us call the attribute objects aX, aY and aZ and their 28

PAGE 29

OIDs xOID, yOID, zOID. The x coordinate value will be wrapped in a data object. An attribute instance will be created and associated with aX using its oid xOID and the data object that wraps the x value will be added to the attribute instance Similarly, attribute instances will be created, one for every other coordinate value. An instance is created with the name Point and all the three attribute instances are added to this instance. Extending this approach to model a complex shape is relatively simple. Consider a box shape that has 8 end points. Each point is represented using the same methodology as above with the exception that instead of creating an instance for each poi nt, an attribute vector called Point is created and attribute instances for each of x, y and z coordinates is added to it. An instance is then created with the name Box and the attribut e vector is added to it. Lyra being an OMS can be used to stor e domain ontologies. Having both the 3D model and domain ontologies stored in a single data st ore makes it possible to tag shapes in the 3D model with semantic information. For instance, consider a 3D scene where a box is used to represent a computer. Assume that we have a domain ontology stored in Lyra that provides taxonomy of computer types as shown in Figure 3.1. If we specify that the box in the 3D scene repr esents a Desktop Computer instance, then the 3D scene automatically gets enriched with the dom ain ontology about computers. Since the box is a desktop computer, it can be inferre d that it is automatically a digital computer. We could use another ontology that lists the average price of digital computer s to get a rough estimate of the price of the desktop computer. Thus, using an OM S helps us combine the 3D visualization aspect with the semantic aspect to produce an ont ology based 3D visualization environment. Connecting to the Lyra database is done using RMI (Remote Method Invocation). RMI provides for remote communication between progra ms written in Java. RMI allows java objects 29

PAGE 30

running on one system to connect to and i nvoke methods on an object running in another machine. A typical RMI server creates the remote objects, makes them available for access and waits for clients to invoke methods on the object s. An RMI client obtains references to the remote object and invokes methods on the object RMI takes care of the communication between the client and the server. An RMI server regist ers or binds its remote object usually with the simple naming facility provided by RMI called the RMI registry. The client looks up the remote object by its name from the RMI registry in th e server and invokes a method on the object. Lyra uses RMI to enable the users to c onnect to a database from anywhere. Computer Analog Computer Digital Computer Node Number Cruncher Server File Server Mainframe Personal Computer Mini Computer Workstation Desktop Computer Portable Computer Figure 3-1 Domain ontology for computers Thus, using Lyra enables us to add structure to the data by organizing the data with well defined domain ontologies and makes re mote access possible through RMI. Addressing the Limitations In the design of landscapes and greenhouses, walk through simulation in a 3-D space is a useful educational tool, especially for understa nding design and for sele ction of alternative designs. Walk through simulations serve to provide a rich user experience as highlighted in [6]. 30

PAGE 31

Storing the entire model in a database a nd operating off the database provides three advantages. The first advantage is that it gives complete control over building the 3D scene. The 3D scene can be modified as it is loaded or after it is loaded dependi ng on the situation. The entire model or parts of it could be pulled up at runtime depending on the necessity. For instance, we could render the 3D greenhous e without the plants and heat ers or with them depending on whether the user is interested in visualizing th e structure of the gree nhouse or inspecting the components inside the greenhouse. If a static VR ML file is used, unle ss we provide a custom VRML browser implementation all modifications to a scene can take place only after the scenegraph is constructed and hence partiall y loading a scene would not be possible. The second advantage is that it simplifies the process of tracking changes to the 3D model. Changes made to the 3D model can be saved by updating the existing model in the database. Updating a 3D model does not pres ent any problems as it would in the case of a static VRML file. Xj3D does not directly suppor t exporting a 3D scene at runtime to a VRML file and even if we altered the SAI implementation to support this, we would have to write it out to a new VRML file. This presents several problems since the user could make multiple changes and saves over a period of time and each save should either save th e entire scene, which impairs user interaction due to disk write delays, or maintain some wa y of incrementally saving the changes to the new file. As a result, there would be multiple copies of the same VRML file. These problems can be avoided using a database because we can change the 3D model and make updates to the database to reflect the changes upon saving and thus having a single instance of the 3D model. The third and the biggest gain from using an OMS to st ore both the 3D model and the associated meta information is that the graphics content and th e semantics of the scenes are combined into a consistent and cohesive ontological model. Users can connect objects in a 3D scene with classes 31

PAGE 32

32 in a domain ontology. By doing this, the 3D scene becomes enriched with semantic information. Associating domain ontologies with 3D shapes also generates a taxonomy which can be navigated using a taxonomy browser such as the Web taxonomy Browser [14].Using an ontology management system also opens up several other po ssibilities as discussed in chapter 5, such as using reasoners to search the semantically augmented 3D greenhouse, positioning the greenhouse on a geo-referenced 2D map and exporting the 3D model to other formats. Sharing models is simplified since anyone can connect to the database remotely. The text based search functionality is sim ilar to the hyperlink based navigation with the difference that the user viewpoint gets reset within the 3D model to the plant of interest. The user can enter search terms and have their viewpoint reset within the 3D model to the region of interest

PAGE 33

CHAPTER 4 ONTOLOGY BASED 3D VIRTUAL GREENHOUSE Domain Ontologies A greenhouse ontology is shown below in figure 4.1. Greenhouse Free Standingstructure Attached Structure Lean To Full Span Window Mounted Quonset T yp e Arched Roof T yp e Gable Roof Type L e ngth Wid t h KW Height E W ma t W W ma t E W ma t W W ma t N W ma t S W ma t N W ma t S W ma t NKW m a t SKW m a t WK W m at EK W m at has3DM odel Ht at m i ddle Figure 4-1 Greenhouse domain ont ology. Grey shaded boxes represent classes. Arrows point from a subclass to its parent. Bl ue shaded boxes are properties. A lean-to greenhouse is a half greenhouse, split along the peak of the roof, or ridge line. Lean-tos are useful where space is limited to a wi dth of approximately seven to twelve feet and they are the least expensive structures. The ridge of the lean-to is attached to a building using one side and an existing doorway, if available. Lean-tos are placed close to available electricity, water and heat. The disadvantages include some limitations on space, sunlight, ventilation, and temperature control. An even-span is a full-size structure that has one gable end attached to another building. It is a costlier option and it provides more space than a lean-to greenhouse. Window mounted structures provid e a convenient method for growing a few plants at relatively 33

PAGE 34

34 low cost. The disadvantage with most of the at tached greenhouses is that if not positioned properly, exposure to the sun may be inhibited by the attached structure. There are also problems with snow or water sliding off from the roof of the attached structure onto the greenhouse and with walls of the attached stru cture retaining the suns heat wh ile the greenhouse cover loses heat rapidly. Freestanding greenhouse structures are separate structures. Th ey can be set apart from other buildings to get more sun and can be made as large or small as desired. The Quonset frame style is a simple and efficient construction with the disadvantage that the sidewall height being lower, storage space and headroom is restricted. As far as structural components are concerned, the Quonset has no side walls compared to the arched and gable style greenhouses. All the three freestanding structures may or may not have kneewalls. The Greenhouse class forms the root of the ta xonomic hierarchy. 3 attributes describe its structure, namely, length, width and height at the center. Attribut e restrictions on each attribute specify that the minimum and maximum cardinality values to be 1 meaning that every greenhouse instance should have these three attr ibutes exactly once. The greenhouse class also has a restriction on an attribute called has3dmodel .This Attribute restriction sets minimum cardinality to 0 for attribute has3dmodel and it is used to link a greenhouse instance with the Lyra object that contains th e 3D rendering information. Freestanding structures may or may not have knee walls. If pres ent, there may be different materials for the north, south, east and west knee walls. This is re flected by creating attributes, one for each knee wall and adding Attribute restricti ons to the Freestanding structures class to set the minimum cardinality to be 0 and maximum to be 1 for each of these attributes. A domain ontology for tomato diseases was created and stor ed in Lyra. [14] was used to create the

PAGE 35

ontology shown in figure 4.2. A 3D object in the virtual greenhouse will be associated with a class type in the domain ontology. By doing this the 3D scene gets transformed into an environment which knows about its 3D objects and can be queried using reasoners to infer facts. Crop Disease Tomato Diseases Tomato BacteriaFigure 4-2 Tomato Plant ontol ogy. Light Gray denotes classes, dark grey denotes instances, and light yellow denotes properties. Database Design for 3D Objects VRML Shape nodes are the only type of nodes fo r which physical entities are created in Lyra. The translation, rotation a nd scale values are stripped off form the parent transform nodes and stored as attributes with the shape node. Attribute vectors are created, one for each shape node present in the VRML file. Each attribute ve ctor records the transl ation, scale and rotation Bacterial Spo t Bacterial Spot on Tomatoes Bacterial Wilt Bacterial Wilt on Tomatoes Tomato Fungi Powdery Mildew Powdery Mildew on Tomatoes Tomato Virus Tomato Spotted Wil t Scientific Name : Xanthomonas campestris pv. vesicatoria Host: Tomato Pathovar: vesicatoria Scientific Name : Ralstonia solanacearum Host: Tomato Scientific Name : Oidium neolycopersici Host: To m ato Scientific Name : Tospovirus Spotted Wilt on Tomatoes Host: Tomato 35

PAGE 36

values inherited from all parent transform nodes along with the coor dinate and texture coordinate values. Each attribute vector is named after the corresponding nodes DEF name if the node has one or a random name based on the nodes position in the hierarchy .All the attribute vectors created are then grouped together and added to the greenhouse instance as an attribute instance of type has3DModel. User Interface The user interface shown in Figure 4.3 is used to construct a greenhouse. The user can specify properties such as th e length, width, height and the te mperature at which the greenhouse must be maintained. By default, the greenhouse is made out of glass. The users can also click on the "expand" button and change the type of struct ural materials for each part of the greenhouse as shown in the Figure. In addition, the user has to specify a VRML file that can be used to visualize the greenhouse in 3D. The user can save the configuration or load an already serialized configuration from the File menu. Methodology For each greenhouse that the user creates, an in stance is created in the database with the user specified name. The instance refers to the one of the differe nt class types specified in the domain ontology as its parent de pending on the greenhouse type that the user chooses. Attribute instances are created to store the length, width ,height, temperature and other properties of the greenhouse and they are added to the greenhouse instance. The VRML parser is run on the VRML file to create attribute vectors that contain the render ing information .The attribute vectors are linked to the greenhouse instance by a dding it as an attribute instance of type has3DModel to the greenhouse instances To view an already serialized greenhouse, the user loads the greenhouse using its name. Lyra is searched for an instance with the user specified name. The instance, if present, is obtained and its attribute instances are extracted. Using the data 36

PAGE 37

present, the user interface is then populated. Using the meta information details extracted from the database, a meta information map is created. Figure 4-3 Swing based UI for setti ng the properties of the greenhouse. The map contains two types of key value pairs. The first type maps a DEF name to a user defined name and the second type maps a user defined name to relevant meta information. This map is kept synchronized with the database a nd any changes made to the map are immediately updated in the database. Mentor Mode Clicking on the "mentor mode" option in the visualize me nu brings up the 3D user interface. The 3D model is re ndered and displayed as shown in 4.4 depending on the greenhouse type the user selects in the user interface.There are several ways users can navigate through this model. Walking enables collision, so if the user collides on a wall or an object while moving he is stopped. The "Fly" mode turns off collision de tection. In addition, users can also pan through the 3D model or tilt the model to adjust their visualization angle. Context clicking on the 3D 37

PAGE 38

model brings up a menu. The users can search the greenhouse or inspect parts of the greenhouse by clicking on the corresponding menu item. Methodology When the user requests an entry into the mentor mode, the 3D model corresponding to the greenhouse the user loaded earlie r is pulled up from the database Using Xj3D, shape nodes are constructed from each of the attribute vectors. A touch sensor node and a viewpoint node are created for each and every shape node. A custom ev ent handler is created and associated with the touch sensor node. The event handler remembers the DEF name of the shape node and provides a call back method that gets fired when a user touches this shape node The call back method creates a dialog that shows meta information a bout the shape node. A tran sform node is created, one for each shape node and the translation, scale or rotation attributes are a dded as fields of the transform node. The shape node and the associated touch sensor and viewpoint nodes are added as children to this transform node. In addition, the sensor node gets added to a vector of sensor nodes and the viewpoint node gets added to a ha shtable which is keyed off the DEF name. The transform nodes are then added to the scenegraph. Inspect Mode When in mentor mode, if the user wants to in spect objects in the 3D model, he/she enters the Inspect Mode. The user can do this by context clicking on the 3D model and selecting the Inspect Mode option. By doing this, all the touc h sensor nodes one per shape node, that were added during the time the scene was loaded,a re enabled by getting each sensor node from the vector and setting their isEnabl ed eventIn to true. The user can switch out of the inspect mode any time by context clicking agai n and selecting the Navigate Mode option. When the user goes back to the navigate mode, the touch sensors are disabled. 38

PAGE 39

Figure 4-4 The 3D greenhouse rendered in Xj3D. In the inspect mode, when the user clicks on a component, a modal dialog is displayed. Information in this dialog is divided by a tabbed pane. The user can enter the user defined name for this component, add viewpoint metrics and textual comments describing this component in the first pane as shown in Figure 4.5.Using the se cond pane, the user can add images relevant to this component. Images that have already been added show up as thumbnail buttons before the Add a new image button. Clicking on the thumbnail renders the full size image. Title for the image can also be changed by clicking on the button next to it. New images can be added to the existing ones on the fly by clicking on the Add a new image button. Clicking on this button brings up a file dialog which can be used to locate the image. Once the image is selected, a thumbnail of the image is created and added to the existing ones. Clicking on the save button saves the new meta information. Figure 4.6 shows images of different types of heaters being added. 39

PAGE 40

Figure 4-5 Textual information pane containing comments, user defined names and viewpoints. Methodology In the inspect mode, when the user clicks on a shape, the meta information map is searched for the shapes DEF name to get the user define d name. The map is searched again based on the user defined name obtained in the previous searc h. This search returns back all the meta details associated with this shape which is then used to fill the meta information dialog with data. When the user clicks on save in the meta info rmation dialog, Lyra is searched for classes or instances with the same SOID as the user defined name. Attribute instances are created and user provided meta information like DEF name, comments and images are added as Data values to the corresponding attribute instances. A new attribute called refersTo is used to link the instance corresponding to user metadata with th e class or instance in the domain ontology. The 40

PAGE 41

Figure 4-6 Image of an electric he ater being added to the 3D object. The image shows the thumbnail and the full sized image. attribute, as its name implies adds semantic in formation by asserting that the 3D object that the user clicked on maps to a particular class type or a particular instan ce in the domain ontology. An attribute instance of type refer sTo is created and it stores the OID of the class or instance in the domain ontology as its data. A ll the attribute instan ces are then grouped together into an attribute vector named after the user defined name and the attr ibute vector is added to the greenhouse instance. Two entries, one with DEF na me and user defined name as the key-value pair and the other with the user defined name and OID of the in stance created previously as the key-value pair, are added to the meta information map. For example, when a user clicks on the 3D obj ect representing a tomato plant, the modal dialog is displayed. The user then tags the object with the user defined name Bacterial Wilt On Tomatoes, types out a short text ual description and a dds some images. When the user clicks on save, Lyra is searched for an object with the SOID Bacterial Wilt On Tomatoes which is an Instance in our domain ontology. An attribute instan ce of type refersTo is created and the OID 41

PAGE 42

of the Bacterial Wilt On Tomatoes instance is adde d as the data value to the attribute instance. Attribute instances are created to store comments and images. All the attribute instances are then grouped together into an attribut e vector and updated back to th e current database instance. Two entries are made in the meta information map ,one being the key value pair and the other being the pair . By this mechanism, the 3D scene gets enrich ed with the semantic information present in the domain ontology. It is possible to use a reasoner in the future as discussed in the next chapter, to make inferences or to process queri es like Show me all plants in the greenhouse that are diseased. The dialog also provides a way by which the user can group different 3D objects into a single entity. For instance, the 3D model for a bench contains several shape nodes one for each face of the bench. This can be verified by clicki ng on different faces of the bench and observing that each face has a different DEF name. Using this tagging func tionality, we can set the user defined name to be the same for all parts of the bench thereby grouping the shape nodes under a common name. Having a common name comes with the additional benefit that meta information added for one shape node of the group shows up when any other member of the group is inspected. Figure 4.7 illustrates this. Searching Clicking on the search menu item brings up an input dialog in which the user enters the user defined name of the component being search ed. Figure 4.8 depicts searching for corn plant. Methodology Searching for a component with its user de fined name requires that the component be tagged with that user defined name. In addi tion, the viewpoint details for the component 42

PAGE 43

A B C Figure 4-7 Aggregating shape nodes under a comm on user defined name. A) Shows the front face of the heater carrying the VRML DEF name Random_child134 and the user name Heater. B) Additions being made to the comments section of the side face that has a different DEF name (Random_Child154) but the same user name. C) Front face reflecting the changes made to the meta informa tion of the side face since they are tagged with the same user name must have been added as a part of meta details. When the user enters a user defined name in the search dialog, the meta information map is sear ched for the name and the viewpoint details for that node are extracted from the data. Then the se t name corresponding to this associated name is found from the map. The hashtable of viewpoint nodes which was crea ted during the loading stage is searched using this set name as th e key and the node relevant to the searched object is 43

PAGE 44

obtained. The viewpoint position in the viewpoint node is set base d on the meta information and its set_bind EventIn is set to true. This results in the view point of the shape node being bound to the scenegraph thereby re-pos itioning the user orientation. Energy Estimation There are two basic equations used to calcu late energy flow through a greenhouse. One is for conduction, the transfer of heat through the structure of the greenhouse, and the second is for convection, the transfer of heat by mixing warm greenhouse air with cooler outside air. When calculating the heat loss for any greenhouse structure, both of th ese equations are taken into account, at least to some degree. The particular equations used are from [13], and are outlined below. To calculate conduction over any greenhouse surface, four values are used; the temperatures on either side of the surface (usually the temperature set point of the greenhouse and the current temperature outside), the area of the surface, and a constant known as the Uvalue of the surface material. The U-value repres ents a typical amount of heat energy that will cross a certain area of the material, per degree difference between the inside and outside temperatures. The amount of heat per hour that is lost to that surface can be found using the equation Q = u A T Equation 4.1 Table 5.1 shows different uValues for differe nt materials that the model supports. For example, if a small greenhouse wa s set to 65F and the outside temperature was 45F, a 100 ft2 sheet of glass with a U-value of 1.13 BTU/ft2 -F-hour, then the amount of heat lost over the sheet of glass would be : (65F 45F) 100 ft2 1.13 BTU/ft2 -F-hour = 2260 BTU/hour. To calculate convection for any partic ular greenhouse, four values ar e used; the temperatures on the inside and outside of the gr eenhouse (the difference being T), the volume of air within the 44

PAGE 45

greenhouse (V), and a correction factor (Cp) representing the relative leakiness of the greenhouse structure. An estimate of heat lost can be obtained using the equation Q = V Cp T Equation 4.2 A B C Figure 4-8 Search demonstration. A) The user labels the 3D pl ant object with the user name corn. B) Shows the user searching for the co rn plant. C) Viewpoint is reset to point to the corn plant For example, if the temperature outside was 45 F, the temperature inside was maintained at 65F, the volume of the greenhouse was 5000 ft3 and the greenhouse was assumed to have an 45

PAGE 46

average 'leakiness' (correction factor = 0.0216 BTU/ft3-F-hour), the amount of heat lost per hour due to convection would be (65F 45F) 5000 ft3 0.0216 BTU/ft3-F-hour = 2160 BTU/hour. Calculations assume average leakiness and al ways use the correction factor value of 0.0216 BTU/ft3-F-hour. To find an estimate of the heat energy lost for the entire structure, conduction is calculated for each wall, each portion of the kn ee wall, and for the roof, and convection is calculated for the whole structure. Adding the resu lts together gives us th e estimate. The values used in the calculation of area/volume of the greenhouse are the structural parameters that the users specify through the interface like length, width and height. A synthetic database of external temperatures, one value per hour over a period of one year, provides the external temperature needed in the calculations. A synt hetic database of expected solar radiation was used to get an estimate of how much heat the greenhouse will ga in from the sun. The amount of light energy hitting per square foot of ground space was multip lied by the total footprin t of the greenhouse to get the estimated energy gain from the sun. Using the above calculations the net amount of heat energy leaving or entering the greenhouse is calculated. This is simply done by adding the total heat lost to conduction and convection to the heat gained by solar radiation. If the result is negative, i.e., the greenhouse is cooling off, the number of BTUs lost is assu med to be the number of BTUs added to the greenhouse by its heaters, to keep it at its curren t temperature. If the number is positive, it is assumed that the greenhouse is be ing cooled somehow and that no fuel is being burned to heat the structure. Comparison Mode The user loads an existing gr eenhouse or specifies the struct ural properties for a new one. Then when a comparison is run with another greenhouse, cost computation is performed for both the greenhouses. A 3D model corresponding to the greenhouse type is pulled up. When the user 46

PAGE 47

clicks on a component, energy lost form that particular component over each time frame is tabulated for both the greenhouses. Table 5-1 Uvalues MATERIAL TYPE U-VALUE GLASS 1.13 PLASTIC 1.15 DOUBLE LAYER PLASTIC 0.70 CONCRETE BLOCK 0.51 POURED CONCRETE 0.75 INSULATED CONCRETE 0.13 Simulation mode Clicking on simulate in the visualization menu loads up a 3D model of the greenhouse. The users can walk into the greenhouse and see th e status indicator on the heater change at run time based on several factors such as the external temperature, the internal temperature or the solar radiation. The indicator goes green when th e heater gets switched on and goes red when the heater gets switched off. In addition the digi tal thermometer on top of the heater updates once every few seconds showing the internal temperature in the greenhouse. It can be observed that the heater gets switched on when the internal temperature in th e greenhouse falls below the set temperature that was specified as a part of th e greenhouse design. The user s can also click on any part of greenhouse and get energy loss st atistics corresponding to that part. Methodology The 3D object representing the light bulb on the heater is tagged with the name Heater. The digital text that shows the internal temperat ure in the scene is a 2D text node. The text node is tagged with the user de fined name Thermometer. During the simulation, at every time step, the inte rnal temperature is calculated for the next time step using the equations described in th e earlier. The 2D text node with the name 47

PAGE 48

48 Thermometer and the shape node with the name Heater are extracted from the scene. The 2D text node is updated with the curr ent internal temperature by calling the set_text eventIn of the text node. Then based on the heaters state, th e Heater nodes color is update by calling the set_color eventIn of the shape node. The extern al temperature data used is from a synthetic dataset that provides external temperatures for every 15 minutes over a period of 24 hours. 2 seconds in the simulation is equivalent to 15 minutes in real world. Figure 4.9 shows the simulation. Figure 4-9 Simulation with the heater showing the internal temperature in the greenhouse.

PAGE 49

CHAPTER 5 CONCLUSIONS AND FUTURE WORK Walk through simulations play a useful ro le in the understanding of design. Previous methodologies adopted in buildi ng a virtual greenhouse focused on either providing just a 3D model or on computer based simulations that cap tured property changes in the greenhouse due to various factors. This research focuses on mee ting both the objectives by coupling the 3D model with its information source. Lyra, an ontology mana gement system, was used as the data store. Xj3D was used as the VRML browser and toolki t. Using Lyra helped in organizing domain specific information by using ontologies and th ese ontologies provided conceptual maps to which the 3D models and other meta informati on sources were attached. Marrying 3D shapes with domain specific ontologies transforms th e static 3D scene which can be used for visualization alone into a semantically rich 3D environment with which multiple users can interact, even from a remote location, by learni ng from existing information sources or adding more meta information. The 3D visualization also improves the search experience by allowing the user to change viewpoint from the curren t location to the destin ation. This creates an experience that is closer to the real world th an navigating static web pages using hyperlinks. A simulation of energy loss from the greenhouse was developed. The 3D visualization enables the user to see the energy loss from each part by clicking on it. Comparison between greenhouses that have the same structure or co mparison of a greenhouse configuration with its older version helps isolat e areas of energy loss. Ontology based search Lyra being an ontology management system s upports defining ontologies as we already discussed. Using Lyra to store the meta info rmation and the 3D models opens up several possibilities related to searching the model. The VR environment can be considered as projecting 49

PAGE 50

50 of objects described by the ontology into a 3D vi sualization environment. The ontology provides a way of describing object properties and behavi ors, describing object taxonomies, and, using ontology reasoners, automatically classifying an d cluster objects into categories. The ontology acts as a dictionary that defines the meaning of concepts, and provides support for natural language references to objects. Ontology reasoners compare the structures of objects to dete rmine how they are the same or different. They can automatically classify new concepts and build categories by grouping together similar objects. This capability leads, among other thi ngs, to query processing, as a query can be classified in order to identify objects related to the query (e.g. find all instances that satisfy the constraints of a que ry). In VR environments, such reasoning processes have many applications. For instance, a user could pose a query such as Show me all free standing greenhouses in this database. The reasoner would infer that Quonset, arched and gabled style greenhouses are all free standing stru ctures. It would then get the in stances of each class type and display a list showing the names of each instance to the user. The user could then select the greenhouse of his/her choice and open up its conf iguration. Another example would be a query like Show me a plant that is diseased. The reasoner sees that there is an instance which refersTo the powdery mildew on tomatoes instance in the domai n ontology. The reasoner infers that powdery mildew is a type of tomato fungi which in turn is a type of tomato disease and it hence classifies this instance as a diseased tomato variety. It gets the user specified name for this diseased instance and loads it up from the database. It gets the viewpoint data and repositions the camera to point to the 3D shape that represents the diseased tomato plant. Natural language Support Support for natural language queries provided by the ontologies enables objects to be named or referenced in many different ways. Multilingual naming and facilities for managing

PAGE 51

synonyms and homonyms enabled the same object to be named and referenced many different ways. The Lyra OMS includes facilities for crea ting dictionaries and storing phrase patterns needed to process more complex definite descriptions in natural language An example of this would be the way the user interface renders the comments and meta information to the user. Based on the user requirements and locality, the UI could leverage the su pport provided by Lyra to store the textual description in multiple languages. Geo referencing the 3D model Most of the virtual tours and virtual greenhouses on the Intern et contain information about where the real world peer is physically located. As illustrated in the literat ure survey, the Virtual Field Day project uses a Flash Object to s how the spatial relationship between different greenhouses, ponds and fields. Support for georef erencing a greenhouse can be done relatively easily. ESRI, a company that specializes in GIS technologies uses Shap e files (.shp) to represent geospatially referenced maps [23]. A parser th at works off the shape file has already been implemented. The parser is similar to the VRML pa rser in its functionality. It works on the shape file and stores the geographic information in Lyra. The data can then be used to generate a flat 2D map of the locality. Th e 3D model can be transformed, scaled and placed at its coordinates. Thus, the user can actually see the georefer enced map and the 3D model placed on the map giving a better viewpoint of the landscape. An example of this is the ISAS project[18] A geospatially referenced campus map was imported into the OMS by parsing ESRI Shape F iles and a detailed 3D model for one of the buildings at the University Of Florida (Rinker Ha ll) was positioned at its coordinates. A similar technique could be used for positioning greenhouses in real environments, such as an existing nursery operation. 51

PAGE 52

52 A B Figure 6.1 Images showing geo-referenced Rinker Hall. A) Front view B) Top View shows the entire map Exporting the 3D model to other formats Storing the 3D model in Lyra makes it possible to export it into a variety of other formats since Lyra provides a clear separation between content and presentation. For instance, the 3D coordinates in the form of attrib ute vectors in Lyra can be expor ted to a VRML file or an X3D file. The exporter would be a software component that operates within Lyra. The target file format is created by the exporter module and this makes it possible to plug -in different exporters to Lyra to output the 3D cont ent in different file formats. Thus, the virtual greenhouse system developed can be expanded to accommodate several such future extensions very easily.

PAGE 53

LIST OF REFERENCES [1] Stacy J. & Shubat, D. & Anna T. (2003).Greenhouse Virtual Tour. University Of Minnesota. http://www.d.umn.edu/biology /greenhouse/VirtualTour/ VirtualTourOfUMDGreenhouse.htm [2] IFAS. (2006). Virtual Field Day. University Of Florida. http://vfd.ifas.ufl.edu/ [3] Tignor, B. et al. (2007). Greenhouse Simulator. University of Arizona. http://ag.arizona.edu/ceac/wge/simulator/greenhouse_content.html [4] Tignor, B. et al. (2007). Worldwide Gr eenhouse Education. University of Arizona, University of Florida, Univ ersity of Vermont, Ohio State. http://www.uvm.edu/wge/index.html [5] Frantz, J. & Buckingham, L.A. & Locke, J.C. & Krause, C.R. (2006). Virtual grower. http://www.ars.usda.gov/services/sof tware/download.htm?softwareid=108 [6] T. Honjo & E. Lim. (200 2). Visualization of Environment by VRML. in Proceedings of the World Congress of Computers in Agriculture and Natural Resources. [7] Aono, M. & Kunii, T. (1984). Botani cal Tree Image Generation. IEEE Computer Graphics & Applications. [8] VRML. (1995).Virtual Reality Modeling Language. World Wide Web Consortium http://www.w3.org/MarkUp/VRML/ [9] ECMA. (1999). ECMA Scripting Language Specification. ECMA -International. http://www.ecma-international.org/ publications/sta ndards/Ecma-262.htm [10] Marrin C. (1997). External Au thoring Interface Sp ecification. SGI. http://graphcomp.com/info/specs/eai.html [11] Xj3D (2006). Xj3D VRML Toolkit Specification. http://www.xj3d.org/ [12] Beck, H.W. (2005). Lyra Ontology Management System. University of Florida. http://orb.at.ufl.edu/ObjectEditor [13] Aldrich & Bartok. (1994). Gree nhouse Engineering, NRAES publication 33 [14] Beck, H.W. (2005). Web Taxonomy Browser. University of Florida. http://orb.ifas.ufl.edu/index.html [15] R. A. Bucklin. (2007). Physical Gr eenhouse Design Considerations. IFAS Extension. University of Florida. http://edis.ifas.ufl.edu/CV254 53

PAGE 54

54 [16] Ross, D.S. (2004). Planning and Buil ding a Greenhouse. University of Maryland Cooperative Extension Service Fact Sheet 645. http://www.wvu.edu/~agexten/h ortcult/greenhou/building.htm [17] R. A. Bucklin. (2002). Florida Green house Design. IFAS Extension. University Of Florida. http://edis.ifas.ufl.edu/AE016 [18] Oliverio,J. & Masako wski, Y.R. & Beck, H.W. & Appuswamy, R. (2007). ISAS: a human-centric digital media interface to empower real-time decision making across distributed systems. In the proceedings of 12th international conference on 3D web technology. [19] Bartok, J.W. (2005). University of Connecticut http://www.umass.edu/umext/floricultu re/fact_sheets/greenhouse_management /jb_building_gh.htm [20] SL. (2007). Second Life. Linden Research. http://secondlife.com/ [21] Smajstrla, A.G. & Zazueta, F.S. (2002). Evaporation Loss During Sprinkler Irrigation. IFAS Extension. University of Florida. http://edis.ifas.ufl.edu/AE048 [22] Java Swing (2007). Suns Java VM. http://java.sun.com/ [23] SHP File Format (1998). ESRI SHP File Format. www.esri.com/library/whitepapers/pdfs/shape file .pdf

PAGE 55

BIOGRAPHICAL SKETCH Raja Appuswamy received his Master of Science in Computer Engineering and Agricultural and Biological Engi neering from University of Fl orida in Fall 2007. He worked with Dr.Howard Beck and professor James Oliver io on ISAS and published a paper titled ISAS: A Human-Centric Digital Media Interface to Empower Real-Time Decision-Making Across Distributed Systems in Proceedings of the twelfth international conference on 3D web technology.He pursed research on Virtual Greenhouses for his Masters thesis under Dr.Howard Beck and Dr.Ray Bucklin. Raja received his Bachelor of Engineering degree in Computer Science from Anna University, India in 2005 55


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E20101117_AAAADE INGEST_TIME 2010-11-18T04:10:13Z PACKAGE UFE0021872_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 551598 DFID F20101117_AACBNO ORIGIN DEPOSITOR PATH appuswamy_r_Page_32.jp2 GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
28a056c06f33de78bc2e09214e1a4b71
SHA-1
43965f29e41cf07ec3457a236b70ee213c27ba53
20375 F20101117_AACBXI appuswamy_r_Page_40.QC.jpg
ea2db29a913dea62f5745e0b354a035c
08ab186ce90540434cf3459e3d84b9c28f4237b3
689 F20101117_AACBSL appuswamy_r_Page_10.txt
bba9b8d0d94803c9b94eb26dd9e6514e
927a15556146fc8dc7f7cf77f5b2f3a699f04587
998707 F20101117_AACBNP appuswamy_r_Page_33.jp2
eb80c11cc1f7900f17951415e4f5663b
0199c3862566aac4633d63b920ac102cc4668568
7121 F20101117_AACBXJ appuswamy_r_Page_42thm.jpg
2e00fb7f0f3cd6992fcdbacfa2a9d3e0
9f94cc2ac4dc6a82aa6061717628e7ef99d20e0b
1051949 F20101117_AACBIS appuswamy_r_Page_15.jp2
f2be363edeaa92546b6183b54cd5181a
04c390499daebfeb2116d46c1632a42a0437789d
2168 F20101117_AACBSM appuswamy_r_Page_11.txt
28e83c461f10edeb055c8b5297287797
adf0cf098c146bd2a110e01ab5b98c8364961695
23256 F20101117_AACBXK appuswamy_r_Page_43.QC.jpg
5ad3cdda3144b6f638be7037e1b3585b
9c4fe82d5a171e73a08b8f59f4bec25bc0a4c189
1279 F20101117_AACBSN appuswamy_r_Page_12.txt
56135c521bb576f6c8cb553605f59416
0fb8633389cc765c76cbae94652e8c7d8df129f6
1051985 F20101117_AACBNQ appuswamy_r_Page_34.jp2
4f1702095b789c08a9b548421a7543b2
1d460cb821ceefbaabc94588c1f09c0e1ccb853e
24601 F20101117_AACBXL appuswamy_r_Page_44.QC.jpg
5a7374611f32a3cc50b873493064c648
0c7375b966a1c9bc2a5af32fef6f6e75013e25c0
4236 F20101117_AACBIT appuswamy_r_Page_08thm.jpg
1cbf6ee626334dfaa503c98b949aa489
ff1c0b026787ec616ea6f24382363155f787fb64
2135 F20101117_AACBSO appuswamy_r_Page_14.txt
9f38b11c161e339b68faabde6fd08853
a1f889ecdc519aa4577380fab4c7e1db4a9b8e0c
1051903 F20101117_AACBNR appuswamy_r_Page_36.jp2
8788849f94bbd870fcc697276284d2e1
3fc12e43c1ef2306eeff5facc90879d9da331104
7220 F20101117_AACBXM appuswamy_r_Page_44thm.jpg
4ccb8187e80ee49d21aba21c62bc3685
37612f336bb453895fbe08d1f3c71cf546797ab2
1826 F20101117_AACBIU appuswamy_r_Page_26.txt
5b63d8ff11b06f20636a5b72e69c33ed
99e49d7d05782e6d7a7699bd8fd0ca3004b2b34a
2162 F20101117_AACBSP appuswamy_r_Page_15.txt
7a176c19ff3bf0daad691c67bbc4ac5d
b83c1ae8d0c1807dd52541c43ac51bb5e2512f63
932739 F20101117_AACBNS appuswamy_r_Page_37.jp2
59bc9d4e8dc31ddb4d7b91c83dec08fa
0af8a5f63e75fc5ebe7193feca8ee7679fc0580d
7389 F20101117_AACBXN appuswamy_r_Page_46thm.jpg
77121b1047801ec0274ee8fb613a018e
bdd7844484392a5b1a5f0c03d6da3b1f144a1265
54394 F20101117_AACBIV appuswamy_r_Page_29.pro
baff9b60f8673370046de636cd8df4a1
e0d95a060167a1faa340d166cd37aafc61e203ea
2209 F20101117_AACBSQ appuswamy_r_Page_16.txt
12b65a6988b7679e67fe8e309c7b3022
bf2224f36e18e8008d2dc449cc817119ca2de3a7
1051981 F20101117_AACBNT appuswamy_r_Page_38.jp2
d8fb325b2fbe6fb7794ea8c61cac5345
4992f1a73dc72931144b0e8cfd27d70c730f4bd0
13315 F20101117_AACBXO appuswamy_r_Page_48.QC.jpg
6b4b4213d1f3c18348f1a831ffaef662
c2138e187f67476cf2b7702281b268ae8dccd5f0
79147 F20101117_AACBIW appuswamy_r_Page_43.jpg
ec69829380ea3b2207b6b95a40ae13fd
cf5f7765b70190470bb81e739df2e2dcf709c9cb
2173 F20101117_AACBSR appuswamy_r_Page_17.txt
d90ba673f01de8061238c16228164a61
32025e1cf81e3008226fb2d0a81cd67a5136d1a9
898682 F20101117_AACBNU appuswamy_r_Page_39.jp2
47dc51f4da3bdbead097d4b87781a3ac
486040338585c18d0e24704e152081dc8e937441
26031 F20101117_AACBXP appuswamy_r_Page_51.QC.jpg
b059b95e3ce899601d8a4facea773704
333b1a8b13834afca83176d9d066b76f4e2e8112
76643 F20101117_AACBIX appuswamy_r_Page_42.jpg
4aefd7a5fc6fb8ffe37ab5e84cb0a417
6aeb9e0a93552898d7dcecd7e4df140b95c100c7
2091 F20101117_AACBSS appuswamy_r_Page_18.txt
dd4670f7bc19dfc2d4df9da7c26c07ac
88b2120bd0cbb551393b1325736f7f16abcca784
1051982 F20101117_AACBNV appuswamy_r_Page_41.jp2
217d0026dc87f9226612ace8793d4722
0660ddfb616c3a7f5ebec5895f65555e462d00bd
4292 F20101117_AACBXQ appuswamy_r_Page_54thm.jpg
260bbc4033c156e231aab80f24044a3b
ae1b953992797ee82333fb33613d1763292389ac
68437 F20101117_AACBIY appuswamy_r_Page_33.jpg
857a4f6dda5de0f248d0ad874ad167be
edceffb29534e267402df4e92a47a877d520578a
319 F20101117_AACBST appuswamy_r_Page_19.txt
c0e8463d69aa59f60b1f27916450ddcd
cc8c165040e9765ffa6111cc295810594ba8aa6b
1051963 F20101117_AACBNW appuswamy_r_Page_42.jp2
bb9785c89b6dee90bf721899a92d395f
cedf28e367dfb96a4db2f91f4ee93334e06c44f4
11136 F20101117_AACBXR appuswamy_r_Page_55.QC.jpg
64d3c3e2849832860f59367b9ec1cdce
a64b3ae5fb8491fed6a76862b419dfe3fa954183
2217 F20101117_AACBIZ appuswamy_r_Page_33.txt
8bb594d21fd48f67400516027da42775
fd207a17f888e160e517c9f7a0013983f5e5754b
2072 F20101117_AACBSU appuswamy_r_Page_20.txt
ca5556834937b31b899799fee4aac3fe
7f7a1654ef0b253603963a2da29b55b1bfeb8546
1051976 F20101117_AACBNX appuswamy_r_Page_43.jp2
ee35718b68168e5c2e33a4334f9cead4
b47561ad0dc51deae9f7d62a387d789f089cec92
2102 F20101117_AACBSV appuswamy_r_Page_21.txt
5535a215b53bc11c5c8a64e8ce8122bb
5707b13957ef1dddac50a62aa399fcc07b3774ea
2004 F20101117_AACBSW appuswamy_r_Page_24.txt
2bc2b92e300652c32c207017bd8228b1
75e2c2d44e8a99fb84b8f2a5c3817bdb79f0e459
84682 F20101117_AACBLA appuswamy_r_Page_11.jpg
4e7005dd03690ac9883dc047f075ab9e
4c6ee92df3d2f8018a9016350fd86d70c701e280
962943 F20101117_AACBNY appuswamy_r_Page_45.jp2
4c0482f7653da51c00393a19ba054d22
0806f3f2689712c562a5c8ba8b203c0e4c7c4ee3
2226 F20101117_AACBSX appuswamy_r_Page_27.txt
933c1162609ad80d9e65ed3b5db56282
9efc2318404ae8977953c6ad7674f88f7a17fc2b
55042 F20101117_AACBLB appuswamy_r_Page_12.jpg
1b4d525660e1294dd2eb1822421f3513
2b94e8909e1df11e111bdbf1df703e4d3ed89727
1051919 F20101117_AACBNZ appuswamy_r_Page_46.jp2
c1d3c3922b2bbfadf1578b4233eabed5
ddf75c0f60cadcbf23bc17ff4d28ea0f3abe6952
2194 F20101117_AACBSY appuswamy_r_Page_28.txt
5b73e62c1b6e04f795356661b9638981
ce648859ad22ce7ce46f51c7719e67dc0904ee9e
79593 F20101117_AACBLC appuswamy_r_Page_13.jpg
552c2ac19520471f6a08cbb844a5e164
53d247fc35fa76753a65b17aad8c6ceef5356fc0
25271604 F20101117_AACBQA appuswamy_r_Page_53.tif
a658e1b17ec884f3ba5af028488f22a7
497ee9cd975d3bd93f85b2ea4c9c395877a29f56
2143 F20101117_AACBSZ appuswamy_r_Page_29.txt
c39acb6d6ed66c51215b2160eef95740
f31296d3f347eca2530af5b958a42e463a2d2623
83583 F20101117_AACBLD appuswamy_r_Page_14.jpg
f16a8770f954d1614e4842b792d78aa3
84839eacd6ba83f3f69c23cd695c34371d60561a
F20101117_AACBQB appuswamy_r_Page_54.tif
bab9346e406a54624d01164192559b7d
776e4c1421634f56ccd158f5fa0651aa26727e83
83559 F20101117_AACBLE appuswamy_r_Page_15.jpg
3f8ae0d10b72511c602ff458116f9cba
da8b0f5b08e763dc030bde6c7b6520ff74dac48f
F20101117_AACBQC appuswamy_r_Page_55.tif
383b8185a80be2e8c7e8922726c4a754
c9a78ecfd087eeb266bd1a865d6e2e5fe04a00b3
79907 F20101117_AACBLF appuswamy_r_Page_16.jpg
431442dac8a37ab46826ce27a5ea95ae
7872e512272f87d76434f7f259c184e1b2380496
7408 F20101117_AACBQD appuswamy_r_Page_01.pro
1a1eaf016a9c41f9e0bb9ba62d91bed2
a9ec020ee32ae2a7dc374392f091f5568739f824
80572 F20101117_AACBLG appuswamy_r_Page_17.jpg
c7607f1a34ab26f8f9b39d9ded5e6a41
b53f429f8481c093a00c04cc9930bf9e0aeed1ce
6185 F20101117_AACBVA appuswamy_r_Page_26thm.jpg
3bd2058571d66e7a0377c91d5af378d7
7994f7e9850946f7facb0b6f654e5bc3ced0d94c
837 F20101117_AACBQE appuswamy_r_Page_02.pro
d3af65c5afd93efb1bcc513a6c22b0ed
f400997875fe059588a9758b66bd7c9911dfa2db
83799 F20101117_AACBLH appuswamy_r_Page_18.jpg
69520ba013eb824d371f48b2e6d9399b
b5a0cbf02dc466e22a68891dfbdf9ccede101ad9
7113 F20101117_AACBVB appuswamy_r_Page_49thm.jpg
fbc68575447f67d775b20c3b6ea6164c
a0b5415c8d1646d89ceb4dd2c0da6ce3f9a4c42b
965 F20101117_AACBQF appuswamy_r_Page_03.pro
28f515e618929f531b388bb2e42917d2
cc21ab4866828027ce0da3d4a814ea9efabfc4b9
77605 F20101117_AACBLI appuswamy_r_Page_20.jpg
435c723cdc7fdecd749e59c1cc4e8b05
9362f17397c70f476822cd8a6cce2bc76e77e276
17734 F20101117_AACBVC appuswamy_r_Page_52.QC.jpg
144f774dcff54005e2f50ca96d935c70
dbaa5ce3185f0e75d7e1abad66f843cce2323641
22035 F20101117_AACBQG appuswamy_r_Page_04.pro
11f5df32f318c678a82c2e33b934addc
f5ede4eb30b70622d8992d566786ba8dfaca88f1
83609 F20101117_AACBLJ appuswamy_r_Page_21.jpg
c6ad92dfe3e3af22fb540f850a12c218
9cb93c99583e090e1c2aae5ce7d660916fc1f5ee
6144 F20101117_AACBVD appuswamy_r_Page_40thm.jpg
497a832675790f577ad5e7e3325ad08b
66403059d5215a9014717a64d9f7154cbfb5d083
102480 F20101117_AACBQH appuswamy_r_Page_05.pro
5ac07f9ee0fde15826fc6ac291b004e1
1f7f88d32708693e18016830d0de7fb867480f82
47083 F20101117_AACBLK appuswamy_r_Page_22.jpg
19a9a74176be4b1ec75478282b2817b6
a8dd8c863feff5f96c9466d0bd7718db56986d67
6738 F20101117_AACBVE appuswamy_r_Page_41thm.jpg
4f42b2112939951a6b36bfff6a67fc38
5d746dcbfeea9ef749757816fcbb0ab76fb94d3b
24438 F20101117_AACBQI appuswamy_r_Page_06.pro
b6e76d4b43a6c22c46afecaf3a1dd4c1
395336714d5c76a6264efc86744bf0b1f605c393
84759 F20101117_AACBLL appuswamy_r_Page_23.jpg
c5b0af8bc655cbfd0d1c50cbb61342d4
4e159c29700f008b3e0600a5206dccc4ba7b19e3
18028 F20101117_AACBVF appuswamy_r_Page_45.QC.jpg
9ee0dc8c4ffd1a9cc56b0322755dee12
2fe5a1549e44382fa268c8e645a3b8dc017369ff
4127 F20101117_AACBQJ appuswamy_r_Page_07.pro
15b561dbf345af0333a62746b1682145
5ea2c51c8106f3c22a37915205fa4354a6610752
82991 F20101117_AACBLM appuswamy_r_Page_24.jpg
9369a143ea10804e5d4f1303e02a2b66
46afadea9d93f35e344a0d3bb1a007e70f13ef69
27974 F20101117_AACBVG appuswamy_r_Page_50.QC.jpg
f35852b1388fe405b73efd26f3bb3a19
398ccafb5e9997ccd68382a283df2e099f1d253c
43552 F20101117_AACBQK appuswamy_r_Page_09.pro
b09c383bb6a67531bb4c20b28ccbc84f
90bab540dc6e2c46783a496182a6963aa37a9558
87188 F20101117_AACBLN appuswamy_r_Page_25.jpg
6a6e4d2800618ade1fb2afa1d0c8dc22
f65498a33525aac6b71ad549c9a776a084ffa47c
20565 F20101117_AACBVH appuswamy_r_Page_05.QC.jpg
c4ccd40fc012a8860dbbe1ef543712b7
9417acaed6ddba164e8a486f100dbeeb96857b9f
17227 F20101117_AACBQL appuswamy_r_Page_10.pro
beb0e5a31849cef7590f9fda0b8c461d
28216e52c7c4fed3ba4628bd10f1b56f5ed08e5a
69792 F20101117_AACBLO appuswamy_r_Page_26.jpg
34bd4937efa4e8a0538f0194f163f7ef
61f1e6eaf8263faa28253589790d161f9f723fd7
3166 F20101117_AACBVI appuswamy_r_Page_03.QC.jpg
3222a5b0d3b973c7ac0eb3f17fe89f93
4edfe8c489fa14a7cdcf625c4768eebcee1c2829
89184 F20101117_AACBLP appuswamy_r_Page_27.jpg
8e5dcd1c7ffc1c531cb80843df39e1e1
901a338a2ccfbc3de6ee52b5a7b73656f9ffedfc
23011 F20101117_AACBVJ appuswamy_r_Page_20.QC.jpg
c7cfdbd0522b3b7a381f2bcb00786c8c
f320fb75de511a106cded4a2299eeca606a3e8d2
53052 F20101117_AACBQM appuswamy_r_Page_11.pro
4e47a9bb1e4ce8edbb8cd38a1430372e
f0975875d3479875ed7f7c91e4a83afe200c743c
84912 F20101117_AACBLQ appuswamy_r_Page_28.jpg
968ccb2d6f132f26036c1ada38b40b82
e728e55298a0e44a1c071361a10e227e0e281baa
7622 F20101117_AACBVK appuswamy_r_Page_25thm.jpg
4aa69108615712f114acdc27cc3b2d4d
fef14c3dea342beb9d8690f8400da8687b6bbe7f
32190 F20101117_AACBQN appuswamy_r_Page_12.pro
09e9067e7bfbf81e1593b1ae1c0e1999
894531bf32e13ed72cef9379e7a2f98de323a438
85237 F20101117_AACBLR appuswamy_r_Page_29.jpg
a74d39b577f9099b713467a4a72fb44a
4bfbf87a1db6cdfe55a2a53eb01619c7320b10c5
26714 F20101117_AACBVL appuswamy_r_Page_28.QC.jpg
761110b5a5874b813c884c698c464128
da137ccadf03310d567444ba7e3acbb202e17445
48096 F20101117_AACBQO appuswamy_r_Page_13.pro
27eee9165e60c21807b9594031316f3d
534f818ebc4b2a3c87df35c8d53f81eab2ed9628
67453 F20101117_AACBLS appuswamy_r_Page_30.jpg
dfacfcf9ee73ad2e3065665ae212150b
12aa7864a29c2ee7fb3b59ab601c17de78abd17a
5084 F20101117_AACBVM appuswamy_r_Page_52thm.jpg
e9d4a3d1c0d7343c4dddc9ad7e1d3589
df70d6c01d174215aca2c83bd8060f6d6217352c
53018 F20101117_AACBQP appuswamy_r_Page_14.pro
7b139554b181e41baa5e674ed097bc1b
1b06ce9c34b756870d539da7fdac442f5d336d48
43526 F20101117_AACBLT appuswamy_r_Page_32.jpg
e7ab72507f28bdc40ed41f12ee24f650
3fce8626cf8a310436b8f16b5346c168be248375
28267 F20101117_AACBVN appuswamy_r_Page_27.QC.jpg
cf468d224deb63c7aa58241f1aca7f93
577873d91ec4f68924d32447296b8af4d5129fea
52049 F20101117_AACBQQ appuswamy_r_Page_15.pro
2c740deae2b8f0b3f66a6fb685dbd24d
e0bdc38b36f84f271f928f44845627b6df1195b7
83702 F20101117_AACBLU appuswamy_r_Page_34.jpg
7f809d6ecbad80a43f8ca3d1f4971d3b
53117f08291ab04dff959030b4ec19cf28008232
6609 F20101117_AACBVO appuswamy_r_Page_16thm.jpg
eff572b383cd6016f7d4511ddcbbae97
d480c2f09c35bf1b810d01272014323d6756b636
51507 F20101117_AACBQR appuswamy_r_Page_16.pro
d69ee87c816b4869f8bc8bbe95a6dc35
6d52af96991e967448cdccfa536bab7428d4e26b
58486 F20101117_AACBLV appuswamy_r_Page_35.jpg
016269e2019ed17ad02d6ed0cf0eb345
1c53bb8bea40f2f1cbf52156da6ffd1d7abe35c8
7156 F20101117_AACBVP appuswamy_r_Page_01.QC.jpg
4a3377f9745515c1b380d63a0f745318
25b4a94dc78b1ca4b4c00ad5a891b7cf66b72b5d
49349 F20101117_AACBQS appuswamy_r_Page_17.pro
a21804c3c0a7972788663d07adc2f6cc
21167ff6f19d4db2982fd01095d1dcaf3defd1e9
24866 F20101117_AACBVQ appuswamy_r_Page_13.QC.jpg
732340631df43da14290965cc2e83e7b
6cd8c05af333c8c2b3eb41eba4ccf75dd1f2f196
52267 F20101117_AACBQT appuswamy_r_Page_18.pro
d762a142a30bcb3412d0e7e23a03d314
b78e4a30e5beac3ca43740927cd503a86a5c3248
81654 F20101117_AACBLW appuswamy_r_Page_36.jpg
a9c844291cefcda913b27fc8391652c9
5ac85e98a3747326947f29ecd5bc2a5e3c6b5d7d
28611 F20101117_AACBVR appuswamy_r_Page_31.QC.jpg
e3e3782b493b68768e33ca8273d67c62
ac4a88302a47e7a7ec44d81700adb62427b71114
3398 F20101117_AACBQU appuswamy_r_Page_19.pro
8dabe96634f14fc3ad3cc45acefb38da
5fcaa1f04fd5bc0f2101f1525286e03a389b9b20
67794 F20101117_AACBLX appuswamy_r_Page_37.jpg
e919196ce2e6edfba80963edfd149bca
2d1b0a7d6eccd9f5885c7e0fcbbe5647c97ae451
25801 F20101117_AACBVS appuswamy_r_Page_38.QC.jpg
fc3aac6245d42b3e9099439a8a5900f5
684003bd286610cba160031e06070541be357b44
48174 F20101117_AACBQV appuswamy_r_Page_20.pro
39328e83215d5afff85167792267bd9d
b78c506f45ad5766ae64e92f8ecb9a6dc0369d9b
80221 F20101117_AACBLY appuswamy_r_Page_38.jpg
48579a679c24b5a3f8fc195b47746a67
542a0d501e0d68ba06d3d8be386a5a981a02b632
5746 F20101117_AACBVT appuswamy_r_Page_33thm.jpg
2a29aba19b4bd4411daf58f6561e8c4e
6bf9e6f46bf04d87887b3e78acfa1057a0431834
52659 F20101117_AACBQW appuswamy_r_Page_21.pro
57f35a0f1eacd5c316360a4acacaebd1
f3ad2c94e7b25e64efbf3af8f5fe20f1d8f959d2
6839 F20101117_AACBJA appuswamy_r_Page_17thm.jpg
63fdedbe43c4d29598ae3d983bcd2b9e
b2ff881977c840757b0cb0c033eb185f0462a5da
61274 F20101117_AACBLZ appuswamy_r_Page_39.jpg
8ddc8bf05810b9e70e5c8aa3aaaca724
a276f923f3511f372fc054809ba05051636628ce
7269 F20101117_AACBVU appuswamy_r_Page_36thm.jpg
21930dc1cd9e5c598e00990700489bdb
662058fe115eaa438ed3e6bd2aefa242773e2af5
26079 F20101117_AACBQX appuswamy_r_Page_22.pro
c4614812158602683c2bd5b3eaeef8f4
8adf1c208e11d71f47d5bdc7dcf0185e8a9ad7e7
1916 F20101117_AACBJB appuswamy_r_Page_08.txt
f88717d58d9dce2270d25901932b656d
de57391dbedf99567219ce3a6237dd86cecefb98
25920 F20101117_AACBVV appuswamy_r_Page_42.QC.jpg
403ee298067c45f84662699d22e9f943
2c48fa22deaf14bbe3bca7966fb483ee9b0913e0
53647 F20101117_AACBQY appuswamy_r_Page_23.pro
fd481b29b8084a5a64bd3efe5f16540a
13583fe550aae95e41c40674b58b285ce06aba12
2124 F20101117_AACBJC appuswamy_r_Page_49.txt
00296733a4fd174ab2ecc91c81deafac
c54eae10648e6e4f311ae3fea023bf610f51b08d
15047 F20101117_AACBVW appuswamy_r_Page_54.QC.jpg
fb9e86b5efd2179cc0b4539dde3cdb57
a3a6000878f9d718b3ad586c0abdba04771560a9
980192 F20101117_AACBOA appuswamy_r_Page_47.jp2
0bdd3f9a2d0dbf1df64ffecbe0087eb8
8ec0685b994ffb14e5ca5aa78c386563f9eb6a51
49954 F20101117_AACBQZ appuswamy_r_Page_24.pro
a43cfb3d973b5131df11abcabc833dc8
4ff94cfe318c6f3b83836f1641498d5cfb5e7ee7
26745 F20101117_AACBJD appuswamy_r_Page_02.jp2
5e478984154320a1e46560a9188c5eb1
16760e441d1446879701cb9cefe2e5eb83037139
21490 F20101117_AACBVX appuswamy_r_Page_30.QC.jpg
c57273338dd8733f31dd13bdd861f05b
ae3adc03a084a53558e6cc96abf38e60712e2520
514055 F20101117_AACBOB appuswamy_r_Page_48.jp2
341b9f4cb33094a53ae53418ace6119e
56bedab9e73fab400f96b253b8ff92830e560f88
854025 F20101117_AACBJE appuswamy_r_Page_52.jp2
14f023bf113a4824e6790760bcd417f0
df50409f25299b9c258c6ed5538be5fa1b95128a
25466 F20101117_AACBVY appuswamy_r_Page_36.QC.jpg
739be2ea95eb0da47cbb95362e508321
0dbfd30687b21aa0233750843f78c21e5db4e979
1051960 F20101117_AACBOC appuswamy_r_Page_49.jp2
d189d1bedc09e7ae8aba9c5bbbcb92dd
e7631596540c69a74cb8075f833a4fa11a871661
46440 F20101117_AACBJF appuswamy_r_Page_08.pro
c553e05f9e88bdcc058a78bd5c1427d2
123fd86dda020e1fae36dd1d92c7971084b5f5de
2031 F20101117_AACBTA appuswamy_r_Page_30.txt
52c5d11a0a8d8c4d49868dd5673e139c
9ee64ee3448eb0a8349717089cc2f69af86cdfce
25151 F20101117_AACBVZ appuswamy_r_Page_17.QC.jpg
9dcd22bcf0e6ee5f6ce6bad6e0bbcf53
2b7f90f6543007a20ea2ec04ccdb6cb748107127
1051977 F20101117_AACBOD appuswamy_r_Page_50.jp2
643d113165e05157a34bf74595e25f6d
75f9774974def1a14908277f1c9802c4a95f5250
F20101117_AACBJG appuswamy_r_Page_23.tif
b3dcc13a73b96e258817ab3369257a69
60f603d9296736d64540ecdaeb2b6360b5a8969e
2240 F20101117_AACBTB appuswamy_r_Page_31.txt
304cbe403c0ec1bfadec639a85cdc934
5d0e200bf0625312a49368425098ad191063e934
1051933 F20101117_AACBOE appuswamy_r_Page_51.jp2
a70587c57926e2ffd860a3377387b2e2
9962d7fdf12b6550c7e32b78af2a78858088d419
89990 F20101117_AACBJH appuswamy_r_Page_07.jp2
e83f05385f1dc535a013113c4ab27ec3
a61800c75f15fa8a146e0302ad3c612df299d719
954 F20101117_AACBTC appuswamy_r_Page_32.txt
d2b48f1c18226795bfad5c405ef9a212
f90f17d9dd9e2b88d855fe2082af049e7e34dd73
1051986 F20101117_AACBOF appuswamy_r_Page_53.jp2
66f6bf5b2d82825c6a8e4465e5316b1a
33fd152c1f35c9f92abf6edb329244b6d3e9e219
F20101117_AACBJI appuswamy_r_Page_11.tif
81b3cdd65407414c28ef0128d016eb1c
6cef119a267821dec1b9ea03d81906d4e906dfa3
2110 F20101117_AACBTD appuswamy_r_Page_34.txt
4c4bb43383d6bfffc25c4780644f05d9
f171954dd41cdbe870d4b879daccc7f792bba80d
918217 F20101117_AACBOG appuswamy_r_Page_54.jp2
dca6e26cfab59e7f86b33a583fced19a
8ff17d0ee782c3a9a4158bf7798c800769d6cec8
F20101117_AACBJJ appuswamy_r_Page_33.tif
8088bbcf4e13a8fc1f875b488f2b0436
fb29e81b7ba09fc90c677fec5c41450e22d6e439
905 F20101117_AACBTE appuswamy_r_Page_35.txt
74309a1c3abd5a85d817dbef3d4e39fa
cbc1bf2377eb25cf79a9767ee56a325052e17615
434773 F20101117_AACBOH appuswamy_r_Page_55.jp2
247fc120021407155695ff3084600921
d417108cac5125d6707d140cffd8fe30cfaa66e3
F20101117_AACBJK appuswamy_r_Page_43.tif
7f5675ef3cff22671e2ed0e714865e9e
65790128dc0f16fcf9b17bd204da69f9800d5f2c
1678 F20101117_AACBTF appuswamy_r_Page_37.txt
bfd311893ed2eb1674a7829e909907a5
ad2eca3d54c0336bddf3c934f1958f0116bda003
F20101117_AACBOI appuswamy_r_Page_01.tif
acc800fbe44c2ffc9c2ed509f610032b
bed5ccf98263b8ad0e85d5efd03705d68bb0f3be
F20101117_AACBJL appuswamy_r_Page_29.tif
79a7a5ca6867031fb08d9c856a2d5b43
42ae6537e0b818ff23670cd274d022e0e71275df
2008 F20101117_AACBTG appuswamy_r_Page_38.txt
20840f35c22cc5dfa8697ea169bb3360
102d4b33497d84ef82935c172d352eb1b5e02253
F20101117_AACBOJ appuswamy_r_Page_02.tif
e150d7012dbf5447978ba3bcdc7ebbe4
ee14ae34ab2fbd1fdf78a907b28062d6129219fa
36017 F20101117_AACBJM appuswamy_r_Page_19.jpg
5daf93cbc1ed26149bb523608e8b86ae
0515960a626f596d0b094c5e81e75604fbacb59a
1186 F20101117_AACBTH appuswamy_r_Page_39.txt
f8bc17c6c57e92e27be2f88d15fe9232
f28f45d4b57d5b52a7c934950634d37c456e6eeb
F20101117_AACBOK appuswamy_r_Page_05.tif
6e6d06230da9b4b02fdd230709e8d27e
f26f80d64a65ee95c98bdc9c0da12f0bcafb7299
2053 F20101117_AACBJN appuswamy_r_Page_51.txt
043d2cccd7ee69e4d3201a7ffaf3bc6f
6ebfeab9a5f66f0279d81df8b0819679876e8578
F20101117_AACBOL appuswamy_r_Page_06.tif
4780d566299a0b88e690d299cfb6010f
802aec7e9492741351b325b0af65369698041db8
990659 F20101117_AACBJO appuswamy_r_Page_40.jp2
8d3f0b53d8d7ecaf6f28d977a74c814d
af0d133d818b625e8795b45d2b38a1676ca661ea
1150 F20101117_AACBTI appuswamy_r_Page_40.txt
16a97e0100c4932879de0c24a318143b
2ba8708e799667ab12f764984bdc8fb5687016e4
F20101117_AACBOM appuswamy_r_Page_07.tif
5fb7b69d886cc1a202020a35feec9c95
73e1dc5c2a1a9c128e2cc0dc5f79331e21b801d7
2014 F20101117_AACBJP appuswamy_r_Page_13.txt
4af2dff6cf22a0f3c7a13622b5748227
c21846b6ea3b185c838de4c28fde9ad73472d344
1434 F20101117_AACBTJ appuswamy_r_Page_41.txt
f736d1862628870f8e5f23db0c6c15ed
8e65ecd9abb4f576023c793bb9fb8d423a9f2df1
F20101117_AACBON appuswamy_r_Page_08.tif
1b45909231e558fbb4cf4371ba16e7b8
012f08ac6f502c6b07b7af1051d403b44f4eba20
1051980 F20101117_AACBJQ appuswamy_r_Page_31.jp2
bbaecb31df1281cbdecbccf91db82034
1afe199e3c8ab32380ec71d0cef3da6f42a9cf0b
1959 F20101117_AACBTK appuswamy_r_Page_42.txt
131cd309f3cc0dbb6e9dea747d2aa993
99dda8c3c6f0c55cb7cb8dfecb6b308cb65e8039
F20101117_AACBOO appuswamy_r_Page_09.tif
219234f1436d1cdfeaf07bb5cdd40153
9b29cc764eee13a77f3403647278d69e0aa094b1
F20101117_AACBJR appuswamy_r_Page_03.tif
98e32cba6f82cf27a4788b3016424ce6
773414e4c9fc90c097e3c7b0a5e2190f18640cb9
2105 F20101117_AACBTL appuswamy_r_Page_44.txt
01e82d2f54bcfd66813f7d34bc305d93
3031e7b7333d9d58130664c74d80c4e645186880
F20101117_AACBOP appuswamy_r_Page_10.tif
9fa8d33814f583790994b83337453f9e
961492be4b794abbaa8866a002544f50631a4a98
7120 F20101117_AACBJS appuswamy_r_Page_11thm.jpg
053fcda0da2d8ac3e0764ed81e829448
2aeac3f0ec5005c032c03ebd9665bf3871d70af6
692 F20101117_AACBTM appuswamy_r_Page_45.txt
4ee3fed8708191b6e744bdeba17c56be
f2381833402e637d66b9eb112e6fc9e112794d7c
F20101117_AACBOQ appuswamy_r_Page_12.tif
47d7a5ddfd685bc22381bf309a5594c6
57ab2b16fe3092e64d7b2bbfabc2690f53e784c8
834811 F20101117_AACBJT appuswamy_r_Page_35.jp2
69998a47cd405dfa82b7769e5c71232b
642ee6952c6999f1a6fc7a5ce3c37234dfe8f6ab
2198 F20101117_AACBTN appuswamy_r_Page_46.txt
cf2b8362dc00f72095630f35fdfb9e88
147eca00d6b5bc8124aaf4c61450d98bd49d6dd5
F20101117_AACBOR appuswamy_r_Page_13.tif
c2ae08aeff609424c83a50d527f24e01
0fd60e6cc9b4464a5b20de58e79047529f444581
1938 F20101117_AACBTO appuswamy_r_Page_47.txt
dcbca33c2a9e57f63ef8f3283abc386d
1410f7ab1619f50b5ccbcb66f99481498d79bd80
F20101117_AACBOS appuswamy_r_Page_14.tif
19544e17fbcdfd410fabe7cbb147e4b3
e4d66e73f318770c1527e5ad2ada94589c094b0d
23542 F20101117_AACBJU appuswamy_r_Page_41.QC.jpg
378dab29f323ee9ee0f4309c81c3020e
68569fb555b3ba398ee38fdf76bdef0790c3c62e
704 F20101117_AACBTP appuswamy_r_Page_48.txt
2ad0e8e60338b8e2bcd9eb538806056e
a2f6bff2fd0decf9757e7a26fae55cdff90d7a0a
F20101117_AACBOT appuswamy_r_Page_15.tif
45f34ee8ef4030078af913cd88dd8563
a1b3bea23f2c5fdf64c155b18cece915e431abf6
F20101117_AACBJV appuswamy_r_Page_44.jp2
cd6a8749da62e2df68f0a106a06d73b8
0993b423be4e5d99ce7ff721ac53d3dcf31f6954
2245 F20101117_AACBTQ appuswamy_r_Page_50.txt
b99d09764a7dc64215bcd6310de0a193
a81b8db3fd78dcde513781503978923f716846dd
F20101117_AACBOU appuswamy_r_Page_16.tif
c9fd32d89d93a1c75966d3e150bc58bf
92b5e5f2890babcb1adfbcf2b22181f31966b953
2106 F20101117_AACBJW appuswamy_r_Page_23.txt
1e0835107f5cf21b7aedb6fbc21ba706
3d2550298e125eab46273d31a19b9b8134582e4d
916 F20101117_AACBTR appuswamy_r_Page_52.txt
a9b6249bb47d47d4319b16070ab9a3b0
41a28fc3e83328f8595b04f52c8d79c38a4f38c5
F20101117_AACBOV appuswamy_r_Page_17.tif
7878d976d9dc90dcfd3f738aa087cfc6
f666f5102d06fc4762dda429ebeb38c470ca5092
27822 F20101117_AACBJX appuswamy_r_Page_46.QC.jpg
cd8a8e734e0809578c7251427d0ed0d5
a362d2a30890e38b585e87b02ee6c73975f7b02c
2088 F20101117_AACBTS appuswamy_r_Page_53.txt
b9b3f59fb52c28773c790a4bc17bcae9
9c28d6938de8c1c8a3d107ec1bc3b501368f07cb
F20101117_AACBOW appuswamy_r_Page_18.tif
9c51ac7e095c89c6d38e6e3bef53a55f
3e0d8b393b7533a81f4bfb8df98752cf3920566c
998481 F20101117_AACBJY appuswamy_r_Page_09.jp2
41ce4d2d84ce34b8518103d34d2a87ef
4089159547effc7272cd0448a71f3f658d117850
1245 F20101117_AACBTT appuswamy_r_Page_54.txt
03b6c4cbb671dcfb22d7bedb1aba39f2
d87fca9268987782b8657898607612e3e47fdcf6
F20101117_AACBOX appuswamy_r_Page_19.tif
81adb3f0b74a6b0bfed985cb3751acf7
4d161ba7036d700c98004be6d4419e80c643f558
3470 F20101117_AACBJZ appuswamy_r_Page_55thm.jpg
3aa890f76b18a69b30b231029bb73ea3
22291384916f221bc9eed0a9e06b95872173bad5
745 F20101117_AACBTU appuswamy_r_Page_55.txt
f5766fe71390be935dc00067a8e04dfd
716e909da6df0dad200e3ffd6e4ac1b6212edf98
F20101117_AACBOY appuswamy_r_Page_20.tif
204f0a5ef3da4eda3859f40bec42859e
959fd9180ae45f997007193968707e6a4f4eb444
2412 F20101117_AACBTV appuswamy_r_Page_01thm.jpg
043af8a145aab88fe7c2857596decb70
0879328ee9608a624e8ded6ba9c1c1078f370953
960563 F20101117_AACBTW appuswamy_r.pdf
87dc67c7fb8564030db389e07e58e2a5
0ecb996b535dc9d90a7a53e80343bfff0f7787e9
62726 F20101117_AACBMA appuswamy_r_Page_40.jpg
de6ade195b76c862022b3d3a3163a574
370d6ccdd273c7ffbae6b3587d6e8a80b259baf3
F20101117_AACBOZ appuswamy_r_Page_21.tif
428b5d18fcc01a5f93298e6523bb0791
b9804655a59c095d30d6e11b857d4435c03ea9a1
7678 F20101117_AACBTX appuswamy_r_Page_31thm.jpg
7ce701af0edadb29f0e5b1345ce2887c
69edddb29d0867a80054d3b2632419743c1b9ce5
72087 F20101117_AACBMB appuswamy_r_Page_41.jpg
615a580b2c6b63acdd7c671076dd675a
ab4e127db7ace9fc8c34c29bc363310678c484fa
15685 F20101117_AACBTY appuswamy_r_Page_08.QC.jpg
0181ccc01167ae35e631be124999dede
d3784873118a07b011330f59a6fe54bfe22184f5
79741 F20101117_AACBMC appuswamy_r_Page_44.jpg
8435b79d77b2e2785d05d99b5eabc58c
a355683692c1074571779a0a3951daf7a7ca0b2f
54663 F20101117_AACBRA appuswamy_r_Page_25.pro
f0e4e70572e6f11a0c51a29ebd5e96b4
e6f9af47476735e7f7fc8c72ebdea279143fc45a
18156 F20101117_AACBTZ appuswamy_r_Page_35.QC.jpg
c4f49f0a003f3a278d63988cfeb19ce8
adb76bd43550efed8554e3e5268a667abe5e247e
56011 F20101117_AACBMD appuswamy_r_Page_45.jpg
895479fd33c00179e0ac1d4c4d7395d7
713794afee943e53f5f83cd0e79a611d815832ad
44348 F20101117_AACBRB appuswamy_r_Page_26.pro
ca25208362003e8810cbe6fce7a02d71
ab4a1376118384d09529775994e9468f9c041e9f
85094 F20101117_AACBME appuswamy_r_Page_46.jpg
39b1769ba093f7c26ef9311b2f1bea0d
5b1bc7bc11e82c26ca596cc3ab14a7e1f17ea682
22721 F20101117_AACBWA appuswamy_r_Page_09.QC.jpg
4fc8aa177bf1beb0acf1c51441fcbc6f
bf199e4be779420b2288d2bdaca87e70929b9e1f
56787 F20101117_AACBRC appuswamy_r_Page_27.pro
44c239d894e7464801f59e64464fe8e6
a6c30926cfcc97324e9358f78d3a2c70eb839dd3
74394 F20101117_AACBMF appuswamy_r_Page_47.jpg
198877dcd40bf0bac7e5a4cd4cd17cd7
470ebeb2a0acd7caa4773aa8fb85c8155d1dfdaa
55785 F20101117_AACBRD appuswamy_r_Page_28.pro
20befd83de5217dd1d7a21eb03a7884e
ac51a3145ebaf5a48ca9c27bc3d2dd084dde996e
40340 F20101117_AACBMG appuswamy_r_Page_48.jpg
96c12dfdface9c5dffb5e42780610064
a69a1b44feeb362582cfe2590ca31b2c68b62049
7001 F20101117_AACBWB appuswamy_r_Page_15thm.jpg
fdc2e0d7f6a65ce77cab76c3490165a2
0eff5ac68f3a42188b5482afc72d967c13cf3baf
39654 F20101117_AACBRE appuswamy_r_Page_30.pro
5c7cdd0b0eacf68633530d0ff0dba35d
8c47fa4509aa56a40e9f57b7826e37cfbe084593
83212 F20101117_AACBMH appuswamy_r_Page_49.jpg
c7d99c27d1ed6ad9ece7838386dea624
ccd7a3ddfb64da93e810cb4f6e2a97de051053f1
6312 F20101117_AACBWC appuswamy_r_Page_09thm.jpg
3c2c3eb05f802eeb758b5f45f3dfee07
df9efe757b8cd9c340cb45c3c54c80e289f6dc8c
57089 F20101117_AACBRF appuswamy_r_Page_31.pro
6b9c0cffcb9d75f3a383cf6dba48e0b6
097f85b8f227bd3f82fedf1afb21c9f97f3da64c
88812 F20101117_AACBMI appuswamy_r_Page_50.jpg
b1e047c4419626dca6914b247244b14d
0b0375a008231926cfbfff4e325b6f51f1c0ec07
21774 F20101117_AACBWD appuswamy_r_Page_37.QC.jpg
c23de0e968451175d50bf58dac81c9fe
8eafb98db9403f1d9dc9d88d7b77320a34525c9f
23866 F20101117_AACBRG appuswamy_r_Page_32.pro
c538af1a2e03dc247502297148cd68c6
3aa885889dbfb76f0d3a2929473e64a2e62b9ecd
80232 F20101117_AACBMJ appuswamy_r_Page_51.jpg
d0561bd31e77a8a8bea327b72645db4c
0c0301f006ff85513c9a78f5ffe0422d04647392
39120 F20101117_AACBRH appuswamy_r_Page_33.pro
4cbd2d7e936c7a762244f07aac32e3f7
f6c4ce00e5575c8cf7b76f041b7932e775eab62a
55359 F20101117_AACBMK appuswamy_r_Page_52.jpg
3eae124948b0aecb623bf986edfdad7f
7c5c9abf1aa0709b4c3b6b92ad685a0724beafef
7597 F20101117_AACBWE appuswamy_r_Page_14thm.jpg
a019c4e8081c330a7e98ce0f54351c91
1ca2b860d9697acc9caae1d99db227ff9969c891
53594 F20101117_AACBRI appuswamy_r_Page_34.pro
69feb92ad2c2dd3c095ef4d68791c37c
2ca000c7c0e3688f2f54d396ff904362bfc24428
85299 F20101117_AACBML appuswamy_r_Page_53.jpg
e55854b0c1818c79f613620c4ed5fd75
b0b72f79204b45668dd26ea6adbfcc81c020b982
7248 F20101117_AACBWF appuswamy_r_Page_28thm.jpg
3a1e2a2d45040a206cfa46a66ed8d140
047d9e0f8ed52251128782baa355ad0f864a6779
22410 F20101117_AACBRJ appuswamy_r_Page_35.pro
0d2c68cae4651cec27cb2a8c6e31c022
abcf89670cd4736ba907c22ff99c3c30741824ae
50028 F20101117_AACBMM appuswamy_r_Page_54.jpg
2a4fe74304ac412d663ef32dfbe18288
ddde9e7e6bdeb9f781fb53a2442d8ce8d5d58853
26435 F20101117_AACBWG appuswamy_r_Page_14.QC.jpg
290ab3d91fdcf44dd966c7686d79cba2
d3c344f9007ba4453f7ae91c5a5cbc3040607fbb
52814 F20101117_AACBRK appuswamy_r_Page_36.pro
76d48ed0193fa7347ecc180de860b7e6
490dc13252e13b6684a64b685da0df9ded65b1e5
36381 F20101117_AACBMN appuswamy_r_Page_55.jpg
d5842ddbff333cc5a165a599613387d5
c5f06f467b056da9aa5467981df8ebf69a6bbd56
3826 F20101117_AACBWH appuswamy_r_Page_04thm.jpg
540bd7d8fed3f03c29a4bef26c69423c
a7988f31c01e50eee5ae0c8f02d4b56ca9ca879e
38976 F20101117_AACBRL appuswamy_r_Page_37.pro
7d7f1518af6a437e9c5b9f9ed08789fd
c2f8ca1ad382c384962e0ea083f8289213021230
227455 F20101117_AACBMO appuswamy_r_Page_01.jp2
4386e0cf187183bd405d700756707893
a8d0628c58749319ca4aa3ae6dfe2a0c653f09d3
86129 F20101117_AACBWI UFE0021872_00001.xml FULL
93ce5d838eb1d1d6f8aeb9581a39d243
872dcf6d799ca80a5fc07a327b512fa530e32c81
49971 F20101117_AACBRM appuswamy_r_Page_38.pro
13f855099ffa91bd676cb516e59af18b
847dec82281a6580eb1fb8c3d73144828fcfcdd1
26867 F20101117_AACBMP appuswamy_r_Page_03.jp2
51df7b77a3fee09ab912d2b753181d1f
75ecaace8d5eef3718923a11e2a1c21bf7f6838f
7675 F20101117_AACBWJ appuswamy_r_Page_06.QC.jpg
c25c79c5a6a56b52c59eee0d84352b38
a462fa509ee823d3299e2142b857bfc54790a080
29160 F20101117_AACBRN appuswamy_r_Page_39.pro
bd97aa65a43ec3f4eaaaddf617058082
be7db32bf5b3666206b78303c230a106f144906b
511566 F20101117_AACBMQ appuswamy_r_Page_04.jp2
88cf8c91d5ec09d12f5fbc42f707aa91
ee46bd941f7ecadaa16b68e8a1100dc7c12d779f
2339 F20101117_AACBWK appuswamy_r_Page_06thm.jpg
7ebe80271ef62fc0807a93b9abe8543f
923de45bf6811c4160e665e5e556a87cada61bb9
28058 F20101117_AACBRO appuswamy_r_Page_40.pro
388ba3edc8d00c638e9fca728a339c77
2bdc86817ffee4820f4237854c03f46b7cdcf27a
F20101117_AACBMR appuswamy_r_Page_05.jp2
9b0a510d6bbd7f4246579aa9d0f3a31f
6e2050b59b328dac639ab736067dac02e9308f54
1585 F20101117_AACBWL appuswamy_r_Page_07thm.jpg
5163c7cdf8d917d1ac3227a07956c33f
e639962853fbe63d0f6831de0ed0a98cd14d20f7
35650 F20101117_AACBRP appuswamy_r_Page_41.pro
d031b6247a2d59a681a9d6cc04f759ed
35e7f86f55dfa35487cc8719f58b5565d3cad97d
603472 F20101117_AACBMS appuswamy_r_Page_06.jp2
7a801ccbcbc4faeb67cdb667a9844c8e
3582d60fd96345b9c5ce9a367d01ede8ef4fa022
10523 F20101117_AACBWM appuswamy_r_Page_10.QC.jpg
1606081df11d9b0df8a0fd3bc6a438d7
47a1200ccf47923bd729d5c537edebfd7ed42271
48267 F20101117_AACBRQ appuswamy_r_Page_42.pro
2fc1176ac213e6c40b942f5849baf4e8
67145d3f71d9ca39d4387bcd0667eab35f62f073
F20101117_AACBMT appuswamy_r_Page_08.jp2
1dec35185d21241ea808a7545f8595dd
66be3df276ba476fe3c8cd82cde9f9066dd9b620
26751 F20101117_AACBWN appuswamy_r_Page_11.QC.jpg
0d3277cda90751b9b357027891946774
b3f3afb20450a69da92faffd8c3005069e6f669f
26769 F20101117_AACBRR appuswamy_r_Page_43.pro
ddfe39b88b894f2c826f98d5dd82ca3c
bf11c6b10d561f679af9c53fd48f9cab3c53515e
391894 F20101117_AACBMU appuswamy_r_Page_10.jp2
1ac0bbccc419dd6681fd0fdcf124b859
633c4cb78a3960f02f180b11ab230d35dbaaf613
17328 F20101117_AACBWO appuswamy_r_Page_12.QC.jpg
a8fb3a08b4489d91cd1a942b141240a5
54b87b9369de3fa1d3bb66c61eec70a61354cc1e
51114 F20101117_AACBRS appuswamy_r_Page_44.pro
ed0eb6674f23b0cf2ab9e237f3c5e85b
cb85bb24fb47abdd2575ca16ed8e17fbd70817b2
1051973 F20101117_AACBMV appuswamy_r_Page_11.jp2
aa03030caac32b4e1404f0dd491a0cb7
8eb309ec934118b30374e9142d157fc4b72afd1c
6957 F20101117_AACBWP appuswamy_r_Page_13thm.jpg
58f74c828f80f7b5e7809ee896a90033
12318550f59eff5427ddf8930b562bd55db1d3cb
16475 F20101117_AACBRT appuswamy_r_Page_45.pro
bfdaa75b0501358ca2a7f423960b60ad
3ef7bfa44b7e4c9c7ea450fc1b33c7de0a85b60d
730082 F20101117_AACBMW appuswamy_r_Page_12.jp2
33988a12262781b3660ca9c8a9dfdfd4
c60544cb79083815294a854f0cdb310aede17c2b
7339 F20101117_AACBWQ appuswamy_r_Page_18thm.jpg
492b00f5717088a10459830424d1d129
4db8ced288a911649b620b724ca158ed4d48965a
54868 F20101117_AACBRU appuswamy_r_Page_46.pro
aa305c11eaa2a604746facaa49c4507e
9c22b15f6bda26a0c73acf954a058542e8c93a56
11333 F20101117_AACBWR appuswamy_r_Page_19.QC.jpg
eb4b7dfd33169eac410ae1efa92e5d75
21d0278a19a64ec3087c825eece2608b3464f7ab
44553 F20101117_AACBRV appuswamy_r_Page_47.pro
7c7a2edab9bf4de5db9afb32fd768f47
8a87c0cc8d2700c0d80343f0168fa427bdf88453
1051961 F20101117_AACBMX appuswamy_r_Page_13.jp2
d9d60c53edbb0c59db956ae58f962dac
e77cba11a686c09f1916055139aa130577691fd9
3421 F20101117_AACBWS appuswamy_r_Page_19thm.jpg
aac52b8ce40964ef1f6e689060926e51
60b7c73c17a6761647233c554a7492042e00b4db
17721 F20101117_AACBRW appuswamy_r_Page_48.pro
851ff291c326fc4064013abfb674a254
cd6a30e18c951fe01d9151e2fedeb4cf113822ba
5881 F20101117_AACBKA appuswamy_r_Page_45thm.jpg
4d6c8f003e668009553b747ae17da2da
84849f5431853023f1812d7ea97c13af0cd8d050
1051971 F20101117_AACBMY appuswamy_r_Page_14.jp2
1d4bcc4faab89c118391fd57f37d3524
8d7515f73417f5f61df97708b95d06853c31a95e
6760 F20101117_AACBWT appuswamy_r_Page_20thm.jpg
40528fe043591f53611a0a205dd6f3a8
78779f13b486790822272115770cc2c8a7ff2831
51214 F20101117_AACBRX appuswamy_r_Page_49.pro
7ad99f4c75402a6829a9b98296e31739
33d47d3a51f033d4b860b025c41c261fbef9af16
6641 F20101117_AACBKB appuswamy_r_Page_53thm.jpg
5712c05dcadd7b7c03ff66244c576d7f
36cce82d591aea3c09acf71ba182c5a735d90b5f
F20101117_AACBMZ appuswamy_r_Page_16.jp2
1ea48a17227648045518ee0f841eb4a2
7b8157cf3bfbfb0015ddf14a261801d0371ddc64
26981 F20101117_AACBWU appuswamy_r_Page_21.QC.jpg
e10023abe120fd2f0756cb8c6cce3019
fa45dd875442accbaa12a1b26a8da24dc2164fc2
56351 F20101117_AACBRY appuswamy_r_Page_50.pro
286512c346c73968489ce6dfcd5cb36f
74a78fefc380243e006c895c1423f41fbe151701
5967 F20101117_AACBKC appuswamy_r_Page_39thm.jpg
d2288c8c84514e2628a4bfab10fa8be6
7e39c703ed9d53cd70513b63334625efb13721b3
7126 F20101117_AACBWV appuswamy_r_Page_21thm.jpg
d9b9f7f7a54733a36a1ed0a8790d8778
5ef3aecc0771c1689e95eecfb403ded3bafcbd54
51263 F20101117_AACBRZ appuswamy_r_Page_51.pro
512926c22e42c45664ccc218cb6cd50c
4e7696904f382556055e4ceed7f6215e6ba8ddf8
2184 F20101117_AACBKD appuswamy_r_Page_25.txt
41572c446a7eebe9fa8be01183541e7e
db72e089d0e42f677c924bf719801767b80f261b
4239 F20101117_AACBWW appuswamy_r_Page_22thm.jpg
e2287bd87e1ae319aa2f32f49c57f680
092a100d40a4a71ccef1432209329c665fde63eb
F20101117_AACBPA appuswamy_r_Page_22.tif
017e32526d663fc57559ca11648e92f3
03177fb8093d980e053673fd7689dddb4576d0c3
F20101117_AACBKE appuswamy_r_Page_48.tif
4933959f8c35c2d933f643d8b67273ee
881e8129cf6554eb35319c91122809fd59e2b294
27097 F20101117_AACBWX appuswamy_r_Page_23.QC.jpg
1cc20d78b84149e883f80fe5b6f22857
fa647ca25e14c077ee36230dd24b1f5106d66863
F20101117_AACBPB appuswamy_r_Page_24.tif
d02ced72199101f5565679e69e06f9ce
606f9e55d34f16f7a12cc964a1117e2219e07dae
1431 F20101117_AACBKF appuswamy_r_Page_43.txt
5723f389d2d6a4870aed3bfcbdae78c7
c57e7972dde3930ef7f22d09ecb87e6d1eedfaee
F20101117_AACBWY appuswamy_r_Page_23thm.jpg
876b7cbbe3e316673927d87d1dc1b25d
6cbdf22f45b1368163dbef2ace4a2331f49c3c71
F20101117_AACBPC appuswamy_r_Page_25.tif
92ca87f625daf748378cf618a41258ae
8d841866e3324a4e213b8a75cacd6fd300734f3c
F20101117_AACBKG appuswamy_r_Page_36.txt
d9914686ac1a5e2edd4ef73a25e9fba2
7dd975dab4e47ab8216d83aeede6832c019ce226
6249 F20101117_AACBUA appuswamy_r_Page_30thm.jpg
3f97d9e9dcde694a5f7c318605243a11
f114d8c86f1d33688ea2832dffc555f720c57f8d
26088 F20101117_AACBWZ appuswamy_r_Page_24.QC.jpg
d4772f9a1b10185c17686cbe2adeaa13
e1990fc2598c90602ea33e8351979a5c101c22da
F20101117_AACBPD appuswamy_r_Page_26.tif
42ecdf2ad017b6e0e3faa0b869a45491
7e49c1ec3991a55ec6b890ce24df5a188b8561ca
27130 F20101117_AACBKH appuswamy_r_Page_18.QC.jpg
7dff2506b8c26126094df8caf56a94fd
a5c4d920e2c42f57be34390e028e0abc1890164f
4894 F20101117_AACBUB appuswamy_r_Page_12thm.jpg
9c9d6f8590ce63417d01c5a2a845c0d2
674d7eb92c2d6298fc7de51dc5a43b6bfb0d3a25
F20101117_AACBPE appuswamy_r_Page_27.tif
27dcd648a0ea09154419b9caaa795b31
fdd0edb450316b785103c7b5be2a8ffdbbba1351
7231 F20101117_AACBKI appuswamy_r_Page_38thm.jpg
6748c0ae9187b2f105fceb47b4dd56cf
34bd6efe0aa57e22281a066dc2f4db5c14dbaec2
22986 F20101117_AACBUC appuswamy_r_Page_26.QC.jpg
3c5f040078c5c58780667c05eea480bb
2cc9f4904933b111589ad46ed5a146c12a7b34e5
F20101117_AACBPF appuswamy_r_Page_28.tif
0c73a98e0756e7ae1c6bc1bc3abf23bd
6edec60e80dcf5cdf03573da78845bc85f7d40da
89560 F20101117_AACBKJ appuswamy_r_Page_31.jpg
86b8eaa0db22cd53f66152ffe4b66d49
e97585f6b24272a0c284c4bc64e8bea2bae0dc54
6507 F20101117_AACBUD appuswamy_r_Page_47thm.jpg
f604b79061dff2ca61c24b38faf49d53
e5d55a20572d79051a24d12add575543043841ac
F20101117_AACBPG appuswamy_r_Page_30.tif
fc77811d955004edaf9311ac9e437a1d
db0de00c7af41c9307732fa07b7afc6eb213c262
173 F20101117_AACBKK appuswamy_r_Page_07.txt
183099c112dfa0a4d33a1a2f6dcda379
61449c2ef1668bf558bed4fc09ed3420186e681f
1334 F20101117_AACBUE appuswamy_r_Page_02thm.jpg
a03c87d49e8fc2f267e3badddf0b307e
83254fecc71d79d1edccc70275ff809ff48e9021
F20101117_AACBPH appuswamy_r_Page_31.tif
1fdc73db987ad92c1467c9066e8ba340
5129218572986b4e870d740500edeadcf40fe07d
1198 F20101117_AACBKL appuswamy_r_Page_22.txt
fcc6c0bae8abb27ddeeed93568fa24b0
b73f2c0d3a443c8e97eda8a5a20943e8699ff54d
3971 F20101117_AACBUF appuswamy_r_Page_32thm.jpg
ef02a2301152c4433cf3aa7c5291f122
0be78b5e2e99b0451aadb3772fc94c3db85eebeb
F20101117_AACBPI appuswamy_r_Page_32.tif
9ac50d824b943c91c08599af2465b8d8
5b337dd7a60f1acfbb36cbcd12cef2befad346df
F20101117_AACBKM appuswamy_r_Page_04.tif
74bbb5ce441f798dd5972829d58bd257
5469e559e1b58db607b1690bd3a9a2a75736fa79
5403 F20101117_AACBUG appuswamy_r_Page_05thm.jpg
15803973194f235d55a9a80e7b0050ec
334c678319d8abf3bf2b7b778216d26bc2fe779b
F20101117_AACBPJ appuswamy_r_Page_34.tif
88afad46c8bee849c29e26389df9165c
81af5852d77aba8c15e615d7534cfbf6acff3584
66849 F20101117_AACBKN UFE0021872_00001.mets
217954f454d762f5aa84e6ac4caa3ec8
d3593099a02267744a7eaec5771bfc9ea0a1bade
7351 F20101117_AACBUH appuswamy_r_Page_27thm.jpg
f8095ead1d0e545f5df404e90b1e2a86
f1d19a58d2b678d2eebef3e6a6457c8ea549b5c8
F20101117_AACBPK appuswamy_r_Page_35.tif
468a1166db0b8e7e0694b3a750051e5e
a5ea6f6caa49ac03bbdbaa3aba45847f7da86b5b
14602 F20101117_AACBUI appuswamy_r_Page_22.QC.jpg
8757eb88fa203a59e4eb7dc90177e9d7
b1ccacd5ed7801f306762076f36a0117e6197bfd
F20101117_AACBPL appuswamy_r_Page_36.tif
ac8cf81791a1e7ea20df108dc971ddbf
7ccf7119c6c85cd459832f0ea19def8dbc18fece
27017 F20101117_AACBUJ appuswamy_r_Page_29.QC.jpg
a2271e0accd6189c976a102751b28729
870810fa542695bf36521a60314d9e12937a2f33
F20101117_AACBPM appuswamy_r_Page_37.tif
8d80efa17fb292cafc72be573ea84abe
0187f6f9dc429ff72b3a5c9d8264da21b6af4bec
23846 F20101117_AACBKQ appuswamy_r_Page_01.jpg
a820eba591a26f38f60c4d3fe4b6e344
fa59534d92dccce423a7441dd45896efb2591722
23530 F20101117_AACBUK appuswamy_r_Page_47.QC.jpg
9a26900bf7f79f1a41d3573521baa8cb
2b2eb82aa10b115d0334c52d1422de516fc88f14
F20101117_AACBPN appuswamy_r_Page_38.tif
f0db7740faf5e7f7c63d60335e6a4c31
0567964fb0eb2b5ceaf8ab62abf1ae008d180e52
10120 F20101117_AACBKR appuswamy_r_Page_02.jpg
9e49e362fe1d54010a03e3ba0d710554
b2d6a3f65e2a6de88c82ef69cb08bb0cf44525ae
12935 F20101117_AACBUL appuswamy_r_Page_04.QC.jpg
fb595346ad0dd93b61aa95ab0849f0f1
3abe1064757519a5e1099d978735af7189228d35
F20101117_AACBPO appuswamy_r_Page_39.tif
c31202d09f7c3620eda2276574bfa641
68c2540f997429e1d8e7587f0d0e6dc787eab1a5
9794 F20101117_AACBKS appuswamy_r_Page_03.jpg
093cc5597acd4aebafbeaecf76d85820
9c2a7ad307df7d6bccf9c387315efd49703bfaf8
3105 F20101117_AACBUM appuswamy_r_Page_02.QC.jpg
b53a93ad9217ee80f6f005ec00e3c600
ca0b735a009928011b084cf5696498ea5f9fab29
F20101117_AACBPP appuswamy_r_Page_40.tif
b06c369ec54921e9a49ce7df2edfd030
0445b27613d91e375eef8c87918b0d3c3f4b7cbd
41228 F20101117_AACBKT appuswamy_r_Page_04.jpg
65eb7a8ed0166c4d9604f9ac3fc6dae9
265951e37fe578564fb254ef3c191c79fd5f8251
6906 F20101117_AACBUN appuswamy_r_Page_51thm.jpg
35f7c3d324692137c0de5da64454f86a
77437d053362a6a7c0872c46f45123ec2c80521a
F20101117_AACBPQ appuswamy_r_Page_41.tif
cb45c8ed251601a76a2babaa34e4f802
f358f2ba85216c5f54969239c421b8774da4855e
26165 F20101117_AACBUO appuswamy_r_Page_34.QC.jpg
005a4795cbb3f625c9c7b4952794496d
92efd1b93937e638aa88aa7b362e2ed914f9fa48
F20101117_AACBPR appuswamy_r_Page_42.tif
20b07c4b88279bfef28b2dffb826cec8
c5502f72e217d6118ef468e9681ffbbea512d1e6
78749 F20101117_AACBKU appuswamy_r_Page_05.jpg
afbc40558742ded3430a4db2829af3bd
335620e385ed89d1cf75cc76acba998f861b232b
7705 F20101117_AACBUP appuswamy_r_Page_50thm.jpg
e51215d276aa457ff33ea279f665791e
c30326801dd56e78ba1e8b92dd612bfdf2a682d5
F20101117_AACBPS appuswamy_r_Page_44.tif
93d4d1ff02826a28a3a7706be19f204a
2ab748c0b11ad5ad4f2d766d11cfc31c48814e9a
3924 F20101117_AACBUQ appuswamy_r_Page_07.QC.jpg
2cab464bfd373adb1f82a032b34edeaf
7d74c990de36ee242f14ff2fe947ade65792a7a9
F20101117_AACBPT appuswamy_r_Page_45.tif
06a4f0af4f716b0a52edf0080b79085e
302e3d75f253e765f0cc7d81875109da13b4ecd4
26239 F20101117_AACBKV appuswamy_r_Page_06.jpg
67a9cbf18d6a5739b8c7f3db58dce80f
af4e4996d02ea5f500d394db00d5332927856ee4
13684 F20101117_AACBUR appuswamy_r_Page_32.QC.jpg
d8728a90ba3dc4265c56161693246a9a
7fb111823be4425be861c4f312194d574170010d
F20101117_AACBPU appuswamy_r_Page_46.tif
3a5e720197ece6bb250650d49a965dc1
e59ce3fa2b69c7e92d39bf9c2b2d9c216b1fc42a
12177 F20101117_AACBKW appuswamy_r_Page_07.jpg
9d04f1c0ad2ee648b75f759a74cb2721
e27d8bdddbce5017f21d7682c8f699c89c1dec83
4294 F20101117_AACBUS appuswamy_r_Page_48thm.jpg
82ed2f0f5b80ddcf42bf5b9c337d6edb
5bb0b47c1d5a60f602bc3a4f4add5f0e853840af
F20101117_AACBPV appuswamy_r_Page_47.tif
097d0578bcb7c023ab0b7115c070183c
3f49f99c9b25c8f6e43527ed3d1d62a9ebbf0de7
52457 F20101117_AACBKX appuswamy_r_Page_08.jpg
bd1a68fd55924d47eac2bc6bfec558be
619344ec511aed28c926646c70938b41033c2b1d
25852 F20101117_AACBUT appuswamy_r_Page_15.QC.jpg
a35847fc9c27a5fbdee48b5ca212dc6c
9090949faab44030e92216e608db13314264d30d
F20101117_AACBPW appuswamy_r_Page_49.tif
35835180333561436d1d643422fb5b47
1096ab6e041c9b227b8a3da8432c5c4b4bf67758
74134 F20101117_AACBKY appuswamy_r_Page_09.jpg
708401cd0048f80f2a60a5d270061f40
c1c3a2b7fea42c9597363af6cf4147b18003cc1b
26378 F20101117_AACBUU appuswamy_r_Page_49.QC.jpg
8d30110868a660c95722799c20856338
b37607ee0c94970b85ded7815fb75b216b140ce1
F20101117_AACBPX appuswamy_r_Page_50.tif
e4c6f6d7dbdb484ec15cfc1376fd1770
87020d0f705799a98c9dde131c3731779f8fb8b5
32169 F20101117_AACBKZ appuswamy_r_Page_10.jpg
22c03eda06ef8d6a055f12f7208a5997
c79b57aba5e642cee951bcf5b2749b4bd9ac8951
23789 F20101117_AACBUV appuswamy_r_Page_16.QC.jpg
e1a8770a81e9aec05c37ab3ea266c921
49db196c49abc19581390a354311e7e80de47205
F20101117_AACBPY appuswamy_r_Page_51.tif
f7446c0cf1182de57c511772f3b67667
cde4bdc2478946bb86293a8a9fc361b441f65329
1363 F20101117_AACBUW appuswamy_r_Page_03thm.jpg
36236a9e7b173bbbdecb3aeee486a58d
a82793bb9c23522151e28afe4a578691dd64d256
1051955 F20101117_AACBNA appuswamy_r_Page_17.jp2
417f0b0470f0a938b94c1087e56954da
fee3c6811d051ae23f402fd0d0cbc5439ec4ba40
F20101117_AACBPZ appuswamy_r_Page_52.tif
3ecb229e7da05cefeaaafd279d29c1e9
3c2b350e7414c27cbd4968940c4379ab9c63f1b5
3186 F20101117_AACBUX appuswamy_r_Page_10thm.jpg
2f26945e00b93976b9148c7f3775fd81
8a97d1d4973f7ebecddb98a40407fcf74c3a48b6
1051948 F20101117_AACBNB appuswamy_r_Page_18.jp2
6964a6c9f6bb68b9957ad1d6753bcff1
76a233c8800e11fd1abb8e2d5b85aefb1b207b13
24668 F20101117_AACBUY appuswamy_r_Page_53.QC.jpg
7ec37affe8bb1ab1509bf728618bd3b7
7a1f63266966fa2489c7697a1c416373121dbac6
771687 F20101117_AACBNC appuswamy_r_Page_19.jp2
eb0a7070377c0be3f5e87064b7e93a51
8509dcc927bf6208213893b29fe7368d9fa5752d
6534 F20101117_AACBUZ appuswamy_r_Page_43thm.jpg
6ac14312d254cd010ae7d80aaafa269e
789ee3dfde21fbfdd231897dd5431ccb42a5bdfe
1051984 F20101117_AACBND appuswamy_r_Page_20.jp2
b3f7b45ea9e9bcf456ff143b28b7c2a4
02497624cc659f6ce8637f9e24f6051efbc1a8fd
21084 F20101117_AACBSA appuswamy_r_Page_52.pro
a4bde46a99795a3456e622ccacac900d
2644889d47b19a48601a22203c99e2cc70ae30be
F20101117_AACBNE appuswamy_r_Page_21.jp2
ee0f6f328c71ed40f973834de6866410
2b77c0e804628af19bc29dd1db5602ba68b93fc1
51134 F20101117_AACBSB appuswamy_r_Page_53.pro
07c912a16d74d2ca5a9c7f9d0b6f5fa1
9599d05a410fb540c508fac94110f308896c4b7a
599176 F20101117_AACBNF appuswamy_r_Page_22.jp2
dc1d23cb65c0d43f39694128ee6c612a
621d5f26064320cc875265c6dfa4544509261242
7036 F20101117_AACBXA appuswamy_r_Page_24thm.jpg
38fe0f13fe423ba90a3ac5e665158166
8e157a506188c393364879a5b2d1647f39c7c6ea
29673 F20101117_AACBSC appuswamy_r_Page_54.pro
6d127665554af671821007b72b22181c
fcffd1fde0d5114311bf27b65fa64f4ad3246a33
1051922 F20101117_AACBNG appuswamy_r_Page_23.jp2
b58214772c87f92038105ff5ca3f21db
02f8ec6f46659952337bff89d6ada30155140d46
27581 F20101117_AACBXB appuswamy_r_Page_25.QC.jpg
557162514b17baa94fa8058dac939885
126e0aedc8778c20f041371dc5cfc6b681909adb
17803 F20101117_AACBSD appuswamy_r_Page_55.pro
863f207712ef4943047c737fdc1958a7
a7bbe223734770e8bcad7aad03ad9ff7a92f43be
1051895 F20101117_AACBNH appuswamy_r_Page_24.jp2
45e5352aaa75c0d2932aeba8fd1bcbe5
7a1a5daa9960fc74e4c12b0c08619e20db540602
423 F20101117_AACBSE appuswamy_r_Page_01.txt
af82a18d346611c04444a82f09b7acc2
705d69b79c09a5baf8dfa9ce89ebce20c418de50
1051943 F20101117_AACBNI appuswamy_r_Page_25.jp2
53a65c163cf7ab0babf44d2a9e450607
33c7adf318af7d8089972946252ba506d2da95d2
7409 F20101117_AACBXC appuswamy_r_Page_29thm.jpg
27dd97df76f5331c0d409fe19be24122
a7016427eb639fc7a7ba33e19a080e84bcaa31f4
88 F20101117_AACBSF appuswamy_r_Page_02.txt
009726b685adc4f0a6318144ecd58460
c9ec35fc012765c823c78aef2b906769cc784372
980739 F20101117_AACBNJ appuswamy_r_Page_26.jp2
5016ec152808db4563726d344bfb043f
104e63bde1a4e16681d3511b8aa051a147df6976
20714 F20101117_AACBXD appuswamy_r_Page_33.QC.jpg
722e58faab07d17752a811359b79529f
2383e2e939a2e51fbaf648d5e1b5390142c25153
79 F20101117_AACBSG appuswamy_r_Page_03.txt
585e2f40f7274268efab929b4217b30d
5b3a9a8548d3d544c7724c2a6533a1a76886f7c8
F20101117_AACBNK appuswamy_r_Page_27.jp2
5ca728e7a99e64f3ad703e4aad0402ab
02c556a745b2050fbf7595ea5b157b8b27c76ab3
7313 F20101117_AACBXE appuswamy_r_Page_34thm.jpg
6acafcc253079af796608752d09dcff0
1ea653d58caf63b533383740cde3638dceb83ca2
932 F20101117_AACBSH appuswamy_r_Page_04.txt
630ce37f5c9deba9669445ba82e16118
95fa801694999b803701bf3ddf57ae0f0d5822b9
1051978 F20101117_AACBNL appuswamy_r_Page_28.jp2
66c338306e317a380378c5be6d798ef1
36dc2d99c02bda035876ab913c7418710d3d483a
5594 F20101117_AACBXF appuswamy_r_Page_35thm.jpg
e492b0125ab76638c66b7760e935ed0c
e236c469ea537c047ffcd2fb5e77c2a06ef852a6
4118 F20101117_AACBSI appuswamy_r_Page_05.txt
dab21203404610de460390a519935437
05529c15d29d279fe5b7d2144559d4d1aa341f57
F20101117_AACBNM appuswamy_r_Page_29.jp2
bee06776e0e2bc566e1e01549934f980
eae4a819567b563bdd55c6f16c000409ec853807
5910 F20101117_AACBXG appuswamy_r_Page_37thm.jpg
d6130a79615d4d525b79dce0858aaa2b
cf5c4194cbd86d00aad2823b8af2f588ede3c922
1099 F20101117_AACBSJ appuswamy_r_Page_06.txt
cd9d7e9bb7e101a1379913ab3e9a4243
4ad37026a8e62a6ab4b54967b15cf753d42e4e80
949817 F20101117_AACBNN appuswamy_r_Page_30.jp2
9bc270e9e1c009aeedfe74788e97fba2
85cb253640c42a86aeacb969a4cd71faac72ba5e
19820 F20101117_AACBXH appuswamy_r_Page_39.QC.jpg
256f7a6bab5ef223a3243de5448d1f21
49499146fc68956e5d93c4aa323a0d0d4bc41e14
1897 F20101117_AACBSK appuswamy_r_Page_09.txt
11a064bdcf39c8c403f6718ecb934d30
a99af5becb8e73c9bd56e2a8f60afa17ca8e4f61