Citation
The Effects of War

Material Information

Title:
The Effects of War Three-Dimensional Comparison Using Historical Aerial Photos of Normandy, France
Creator:
Balough, Nathaniel David
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (91 p.)

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Forest Resources and Conservation
Committee Chair:
Wilkinson,Benjamin E
Committee Co-Chair:
Barnes,Grenville
Committee Members:
Abd-Elrahman,Amr H
Graduation Date:
5/1/2020

Subjects

Subjects / Keywords:
aerial -- normandy -- photogrammetry -- photography
Forest Resources and Conservation -- Dissertations, Academic -- UF
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Forest Resources and Conservation thesis, M.S.

Notes

Abstract:
Analog single-exposure historical film negatives may seem antiquated in a digitally focused world, but analog images provide a great example of how the past can help the future. Normandy was the largest amphibious assault ever, conducted by the United States Army and British troops, and required detailed planning, airborne reconnaissance, naval artillery, and ground and airborne troops for the assault. The historical Normandy film imagery at the time was all analog, but through current digital photogrammetry processes, these images come alive and show the areas the soldiers fought through with a detailed visualization, and also how the landscape changed because of the invasion. After conducting image retrieval, image conditioning, and photogrammetric processing, the images were used to produce an orthorectified mosaic with a resolution of 41.2 cm/pix and a dense 3D point cloud. The measured absolute horizontal, absolute vertical, and relative height root mean square errors (RMSEs) were found to be 4.21m, 2.92m, and 3.85 m respectively when compared to DigitalGlobe products, and 3.85 m, 4.82 m, and 4.00 m respectively when compared to Google Earth products. DigitalGlobe and Google Earth both use the World Geodetic System 1984 (WGS 84) for horizontal and vertical data and the Shuttle Radar Topography Mission-2 (SRTM-2) data for height. Agreement between the historically derived products and modern sources is evidenced by a calculated cliff height of 60.06 m compared to the DigitalGlobe and Google Earth-measured 60 m height. The historical products also showed promise for assessing battle damage in the form of bomb-crater volume calculations derived from them. ( en )
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.
Thesis:
Thesis (M.S.)--University of Florida, 2020.
Local:
Adviser: Wilkinson,Benjamin E.
Local:
Co-adviser: Barnes,Grenville.
Statement of Responsibility:
by Nathaniel David Balough.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
LD1780 2020 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

1 Permission to Reproduce Copyrighted Material Any candidate who intends to quote or reproduce material beyond the limits of "fair use" from a copyrighted source must have written permission from the copyright holder. A copy of this written approval must be submitted to the Graduate School Editorial Office no later than the final submission date of the term the candidate graduates. The form below is intended to aid the candidate in fulfilling his responsibility. PERMISSIO N TO REPRODUCE COPYRIGHTED MATERIAL We , Encyclopdia Britannica, owners(s) of the copyright of the work known as Normandy Invasion Overview , hereby authorize Nathaniel Balough to use the following material as part of his thesis to be submitted to the University of Florida. Website https://www.britannica.com/event/Normandy Invasion/images videos Image to be Reproduced Map of crossing routes during the Normandy Invasion _______________________________________ Signature of Copyright Holder _______________________________________ Date



PAGE 1

1 Permission to Reproduce Copyrighted Material Any candidate who intends to quote or reproduce material beyond the limits of "fair use" from a copyrighted source must have written permission from the copyright holder. A copy of this written approval must be submitted to the Graduate School Editorial Office no later than the final submission date of the term the candidate graduates. The form below is intended to aid the candidate in fulfilling his responsibility. PERMISSIO N TO REPRODUCE COPYRIGHTED MATERIAL I , Dirk Spenne mann , owner of the copyright of the work known as Fairchild K 17A Aerial Camera (sn# 4484) n 11 hereby authorize Nathaniel Balough to use the follo wing material as part of his thesis to be submitted to the Univers ity of Florida. Website https://www.flickr.com/photos/heritagefutures/6361975269/in/pool camerawiki Image to be Reproduced Fairchild K 17A Aerial Camera _______________________________________ Signature of Copyright Holder _______________________________________ Date



PAGE 1

THE EFFECTS OF WAR: THREE DIMENSIONAL COMPARIS ON USING HISTORICAL AERIAL PHOTOS OF NORMANDY, FRANCE By NATHANIEL D. BALOUGH 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 2020

PAGE 2

© 2020 Nathaniel D. Balough

PAGE 3

To Morgan for her unconditional support, Timothy Kolakowski, John Balough, Harry Swinehart, and all those that have gi ven their lives for our country

PAGE 4

4 ACKNOWLEDGMENTS I thank the chair and members of my supervisory committee for their mentoring, the staff and members at the UF Libraries and the National Archives and Records Administration for their keen research assistance. I thank my wife, children, and parents for their loving encouragement, which motivated me to complete my study.

PAGE 5

5 TABLE OF CONTEN TS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS ................................ ................................ ........................... 10 ABSTRACT ................................ ................................ ................................ ................... 11 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 13 Objective ................................ ................................ ................................ ................. 15 World War II ................................ ................................ ................................ ............ 16 Operations Sledgehammer and Roundup ................................ ........................ 17 Operation Neptune ................................ ................................ ........................... 18 Historical Imagery ................................ ................................ ................................ ... 19 Vertical and Oblique Photography ................................ ................................ .... 20 Fairchild K 17 Camera ................................ ................................ ...................... 23 Military entities ................................ ................................ ................................ .. 25 Naval vessels ................................ ................................ ............................. 25 Aircraft units ................................ ................................ ............................... 25 2 METHODS ................................ ................................ ................................ .............. 37 Image Retrieval ................................ ................................ ................................ ....... 37 Initial Overlay Identification ................................ ................................ ............... 38 Overla y Beach Identification ................................ ................................ ............. 39 Scan of Identified Overlays / Images ................................ ................................ 39 Preprocessing ................................ ................................ ................................ ......... 40 Image Conversion ................................ ................................ ............................ 40 Fiducial Measurement ................................ ................................ ...................... 41 Projective Transformation ................................ ................................ ................. 41 Artifact and Noise Removal ................................ ................................ .............. 43 USGS Camera Calibration Reports ................................ ................................ .. 44 Photogrammetric Processing ................................ ................................ .................. 45 Image Alig nment ................................ ................................ .............................. 45 Optimization ................................ ................................ ................................ ..... 47 Photogrammetric Products ................................ ................................ ............... 48 Dense Point Cloud Analysis ................................ ................................ ............. 48 3 ANALYSIS AND RESULTS ................................ ................................ .................... 57

PAGE 6

6 Overlays and I mages ................................ ................................ .............................. 57 Preprocessing ................................ ................................ ................................ ......... 59 Bandpass Filter ................................ ................................ ................................ 59 Noise Removal ................................ ................................ ................................ . 60 Photogrammetric Processing ................................ ................................ .................. 60 Results and Analysis ................................ ................................ ............................... 61 Control / Checkpoints ................................ ................................ ....................... 62 Crater / Cliff Size ................................ ................................ .............................. 63 4 CONCLUSION ................................ ................................ ................................ ........ 8 0 APPENDIX: OPERATION NEPTUNE OPERATIONS ORDER ................................ ..... 82 LIST OF REFERENCES ................................ ................................ ............................... 87 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 91

PAGE 7

7 LIST OF TABLES Table page 1 1 Table of Combat Squadrons by aircraft flown during the combat missions flown over the beaches near Normandy ................................ ............................. 27 2 1 Example of the format used for locating overlays and images to scan for processing ................................ ................................ ................................ .......... 56 3 1 Number of overlays ................................ ................................ ............................ 65 3 2 Ph otogrammetric processing data point values ................................ .................. 65 3 3 Results of the check point analysis of DigitalGlobe and Google Earth Pro ......... 65

PAGE 8

8 LIST OF FIGURES Figure page 1 1 The invasion beaches and the crossing routes across the English Channel ...... 28 1 2 Aerial imagery day of invasion with propeller ................................ ...................... 29 1 3 Viewfinder used by K 17 camera to ensure required overlap ............................. 30 1 4 Aerial image with fiducial marks ................................ ................................ ......... 31 1 5 K 17 camera ................................ ................................ ................................ ....... 32 1 6 Cross section of K 17 camera ................................ ................................ ............ 32 1 7 Image of the P 38 Lightning ................................ ................................ ............... 33 1 8 Image of the P 51 Mustang ................................ ................................ ................ 34 1 9 Imag e of the F 3 Havoc ................................ ................................ ...................... 35 1 10 Image of the B 17 Flying Fortress ................................ ................................ ...... 36 2 1 Overall Workflow for project ................................ ................................ ................ 49 2 2 NARA Workflow for project ................................ ................................ ................. 49 2 3 49N0 00W Map ................................ ................................ ................................ .... 50 2 4 Over lay 11 ................................ ................................ ................................ .......... 51 2 5 Overlay 11 superimposed on map ................................ ................................ ...... 52 2 6 Example of light lines in imagery ................................ ................................ ........ 53 2 7 Google Earth point locations ................................ ................................ .............. 54 2 8 USGS PhotoScan workflow ................................ ................................ ................ 55 3 1 Pre invasion orthomosaic ................................ ................................ ................... 66 3 2 Pre invasion orthomosaic east of Figure 3 1 ................................ ...................... 67 3 3 Post invasion orthomosaic ................................ ................................ .................. 68 3 4 Pre invasion filtered images ................................ ................................ ............... 69 3 5 Post invasion filtered images ................................ ................................ .............. 70

PAGE 9

9 3 6 space error ................................ ............................ 71 3 7 Dense point cloud s ................................ ................................ ............................. 72 3 8 Post invasion orthomosaic overlaid on current Normandy area imagery ............ 73 3 9 Post invasion orthomosaic overlaid on current Normandy area imagery with control and check points shown ................................ ................................ .......... 73 3 10 Cloud Compare manipulated point cloud to show cliff height calculation ........... 74 3 11 Raw and processed image of the crater and battle fortification .......................... 74 3 12 Current Google Earth view of the crater selected ................................ ............... 75 3 13 Overview of area with inset of crater ................................ ................................ .. 75 3 14 Grenville Barnes. Normandy crater . March 03, 2020. Longues sur Mer. ............ 76 3 15 Grenville Barnes. Normandy battle fortification . March 03, 2020. Longues sur Mer. ................................ ................................ ................................ .............. 77 3 16 Volume calculation of crater ................................ ................................ ............... 78 3 17 Volume calculated grid inset in crater ................................ ................................ . 79 A 1 Operation Neptune operations order page 1 ................................ ...................... 82 A 2 Operation Neptune operations order page 2 ................................ ...................... 83 A 3 Operation Neptune operations order page 3 ................................ ...................... 84 A 4 Operation Neptune operations order page 4 ................................ ...................... 85 A 5 Operation Neptune operations order page 5 ................................ ...................... 86

PAGE 10

10 LIST OF ABBREVIATIONS APFO Aerial Photography Field Office BMP Bitmap D EM Digital Elevation Model DIA Defense Intelligence Agency DOD Department of Defense DOQ Digital orthophoto quadrangles GBR Great Britain, UK GIS Geospatial Information Systems GSD Ground Sample Distance NARA National Archives and Records Administration NCAP National Collection of Aerial Photography QGIS Quantum Geospatial Information System RG Record Group RMSE Root mean square error SfM Structure from Motion SRTM Shuttle Radar Topography Mission TIFF Tagged image file format U K United Kingdom USDA United States Department of Agriculture USGS United States Geological Survey WGS World Geodetic System WWII World War II

PAGE 11

11 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 THE EFFECTS OF WAR: THREE DIMENSIONAL COMPARISON USING HISTORICAL AERIAL PHOTOS OF NORMANDY, FRAN CE By Nathaniel D. Balough M a y 2020 Chair: Benjamin Wilkinson Major: Forest Resources and Conservation Analog single exposure historical film negatives may seem antiquated in a digitally focused world, but analog images provide a great example of how the past can help the future. Normandy was the largest amphibious assault ever, conducted by the United States Army and Br itish troops, and required detailed planning, airborne reconnaissance, naval artillery, and ground and airborne troops for the assault . The historical Normandy film imagery at the time was all analog , but through current digital photogrammetry processes, t hese images come alive and show the areas the s oldiers fought through with a detailed visualization, and also how the landscape changed because of the invasion. After conducting image retrieval, image conditioning, and photogrammetric processing, t he images were used to produce a n ortho rectified mosaic with a resolution of 41.2 cm/pix and a dense 3D point cloud . T he measured absolute horizontal, absolute vertical, and relative height root mean square errors ( RMSEs ) were found to be 4.21 m , 2.92 m , and 3.85 m respectively when compared to DigitalGlobe products , and 3.85 m , 4.82 m , and 4.00 m respectively when compared to Google Earth products . DigitalGlobe and Google Earth both use the World Geodetic System 1984 (WGS 84) for horizontal and vertical data and the Shuttle Radar Topography Mission 2

PAGE 12

12 (SRTM 2 ) data for height. Agreement between the historically derived products and modern sources is evidenced by a calculated cliff height of 60.06 m compared to the D igital G lobe and G oogle E arth measured 6 0 m height. The historical products also showed promise for assessing battle damage in the form of bomb crater volume calculations derived from them.

PAGE 13

13 CHAPTER 1 INTRODUCTION Photography for mapping was first collected usi ng terrestrial methods. In 1860 , after Albrecht Meydenbauer almost fell down a cathedral wall , he wrote a memorandum about the documentation of buildings through photography , and he described that photographic images can store the object information in great detail and with high accuracy. 1 Following the invention of the airplane in 1903 , aerial photos have been continuously captured for mapping large areas around the world for a diverse swath of applications from agriculture to military reconnaissance. A erial photography has been collected by numerous agencies within the US government for a variety of purposes for over 100 years, including World War s I and II, and often times for the purposes of mapping . Similarly, many different agencies maintain histori cal records of the aerial photography captured . The United States Geological Survey (USGS) maintains a variety of geospatial data of the US. This includes digital orthophoto quadrangles (DOQs) captured from aircraft and, more recently, drones. A DOQ is a c omputer generated image of an aerial photograph in which the image displacement caused by terrain relief and camera tilt has been removed. The DOQ combines the image characteristics of the original photograph with the georeferenced qualities of a map. 2 T he United States Department of Agriculture (USDA) maintains a record of state aerial photographs from federally contracted flights supporting geospatial information for land disputes, agricultural change detection, or environmental assessments. The National Archives and Re cords Administration (NARA) maintains a large assortment of historical documents for the government , including publicly available wartime aerial photos . 1 (Albertz, 2002) 2 (Earth Resources Observation and Science (EROS) Center)

PAGE 14

14 The fact that different agencies maintain and hous e similar type s and coverage area s of aerial images makes it challenging to identify the availability of the data and obtain it due to inconsistencies in data preparation, indexing techniques , and met adata formats . For example, s ome federal agencies p rovide only unprocessed historical aerial photography , while others provide them orthorectified. The Aerial Photography Field Office (APFO) , a division of the USDA, is home to more than 70,000 film rolls of imagery (over 10 million images), covering most of the United States and its territories, and did not start to orthorectify their imagery until the early 2000s . 3 The National Archives and Records Administration in Washington D.C. contain s many important documents including the Declaration of Independence, the Constitution, and the Bill of Right s, but NARA maintains records at many locations across the country. NARA s goal is to preserve records tracing the history and story of America for both the government and the American people through a nationwide network of cen ters, p residential libraries, and online databases . 4 The National Archives maintains a r ecord of historical war imagery. Dep ending on the type and scope of the specific imagery , it may be housed at one of a number of different national facilities, libraries, or research centers. The U nited Kingdom (UK) also maintains a similar aerial repository. Their main collectio n is in Edinburgh, Scotland, and they started a digitization partnership with NARA in 2017. NCAP is the National Collection of Aerial Photography. It is one of the largest collections of aerial imagery in the world, containing tens of millions of aerial images featuring historic events and places around the world. It is the role of NCAP to collect and secure the future of these 3 (United States Department of Agriculture) 4 (National Archives and Records Administration)

PAGE 15

15 records, both digital and physical, to preserve them for generations to come, and to make them as accessible and available as possible. 5 Many agencies have digital copies for purchase, or allow individuals to digitally scan analog copies for free. This second method of visiting the agency and scanning allows for consistency in image acquisition. Objective The objective of this research is to augment basic change detection to determine event analysis or battle d amage assessment from historical photos using modern digital photogrammetric processes without traditional surveying techniques. Th e goal is to generate methods appli cable to historical imagery of hostile or inaccessible areas to assist in assessments of blast s or damages, as well as general change detection in three dimension s . The Allied invasio n of Normandy ( France ) was a crucial battle during World War II (WWII) , and the invasion regions included many cliffs and heavily fortified and hostile areas. T his event produced a large amount of various types of historical aerial imagery of a small geogr aphic area before, during , and after significant changes occurred. This research employs that imagery to inform the development of photogrammetric workflows, with the above goals in mind, using contemporary photogrammetric software. Historical images, whi ch can be challenging to obtain, are important in determining or detecting the effects of war . P hotogrammetric ally p rocessing historical imagery create s georeferenced layer s to compare timeframes before and after the invasion . The imagery can help understa nd the environment, the layout of the beaches 5 (National Collection of Aerial Photography)

PAGE 16

16 an d surrounding area, and how the war may have affected the residents near the invasion areas . This research highlight s the differences between historical film and modern remote sensing technology . World War II World War II started in 19 39, lasted about six years, and was the largest and most violent armed conflict in the history of mankind. 6 WW II pitted the Axis P owers (Germany, Italy, and Japan) ag ainst the Allies ( UK, France, and Poland). The war began whe n Germany invaded Poland in September 1939, which precipitated Great Britain and France decl aring war on Germany days later. In early 1940, Germany subsequently invaded many Western European countries, including the Netherlands, Belgium, Luxembourg, and ev entually France. Later in 1940, Germany expanded n orth and e ast invading Norway, Denmark, and Romania. President Franklin D. Roo sevelt stated the US was averse to joining WWII, but still sign ed the Atlantic Charter with the UK Prime Minister , Winston Chu rchill , in 194 1. The Atlantic Charter was a US pledge to the UK to ensure the defeat of the Nazis and the Axis Powers. President Roosevelt and Prime Minister Churchill discussed when and where the priorities of the US would be when t hey got involved with the war. They decided on a two echelon philosophy for the Atlantic Charter ; one tier (theater) focused on the F irst and the other tier was a F irst (country) strategy as First rmany First strategie s established battlefield priorities for the Allied forces over other Axis powers and locations ( Pacific and Japan). 6 (Hammond, 2003)

PAGE 17

17 During the initial years of WWII, Adolf Hitler focus ed on the Eastern F ront, due to the ability of his al to ex pan d / inva de and fight against more powerful countries and armies including the Soviet Union . Germany had already invad ed most of t he Western countries (Western front) and the invaded countries (France, the Netherlands, Belgium, and Luxembourg) had no strong allies to assist them except Great Britain . On 7 December 1941, Japan stealthi ly attacked the US at Pearl Harbor, and within 24 hours President Franklin D. Roosevelt declared war on Japan. T he other Axis Powers declared war on the US within d ays. As the war progressed, President Roosevelt tasked General Dwight D. Eisenhower , US Supreme Commander Allied Expe ditionary Force , with two separate potential plans for the N orthern European T heater ; one was a possibl e inva sion (Operation Roundup) , or a buildup of forces in Great Britain (Operation Sledgehammer). Operations Sledgehammer and Roundup The planning operations that General Eisenhower conducted were in concert with the national level strategic plans of President Roosevelt, Prime Minister Chur chill, and Soviet Union Premier Joseph Stalin. Operation Roundup was an invasion into France in 1943, to alleviate pressure on the Soviet Union by creating a Wes tern Front, forcing Germany to fight on both the Eastern and Western F ronts. Operation Sledgeha mmer was a planned invasion in 1944, after a build up of Allied forces in Great Britain . Operation Sledgehammer later d eveloped into Operation Roundhammer and Operation Overlo r d which included the invasion of the beaches at Normandy (Operation Neptune), th e area of interest for this study. Appendix A is a copy of the operations order issued for the airborne troops in Operation Neptune. The strat egic operation in the Southern T heater just prior to the invasion at Normandy was Operation Torch and

PAGE 18

18 included fig ht ing in Northern Africa against General m ajor (equivalent to M ajor G eneral in the US military) Rommel (Desert Fox) , a very successful German military leader. The battles against Rommel caused delays in the Southern T heater due to his military prowess and leadership and subsequently affected the N orthern Theater operations. These delays ultimately led to the decision to select Operation Sledgehammer over Operation Roundup. Operation Neptune Normandy , D Day , and Operation Neptune are all names for the same i nvasion event that occurred on 06 June 1944. The invasion involved US , Canadian , and British forces across five separate beaches ( code named Utah , Omaha , Juno , Gold , and Sword ). Figure 1 1 illustrates the beaches for the invasion and the crossing routes ac ross the English Channel . was the bloodiest of the D Day beaches, with roughly 2,400 U.S. troops turning up 7 The planning of the time and place of the Normandy invasion was very important to the success of the operation . General Eisenhower knew the bea ches at Normandy would be a n effective position to invade France because of s proximity to France acros s the channel. The biggest unknown for the Axis Powers was whether the Allies would utilize Normandy or the much closer port of Calais as the infiltration point . The weather also played a factor in planning, and it was an advantage for the Allied Forces. T he only factor that could have favored the Germans on 6 June was the weather: there was a heavy cloud cover over Normandy that morning which would have permitted movement of the two panzer divisions without risk of serious interdiction from the Allied air forces. By the time they were 7 (Greenspan, 2018)

PAGE 19

19 given permission to move the cloud cover had broken and nothing could be done before dark. 8 The Allies began document ing their aerial reconnaissanc e in 1941 near Normandy, three years before Operation Neptune , and continued un ti l over a year after . The US started increased aerial reconnaissance in the middle of 1943 and continued their aerial reconnaissance of this area of France for at least a year after the invasion. Historical Imagery This research demonstrates how imagery played a significant role in the success of D Day. Additionally, it provides proof of concept that the application of modern photogrammetric software can enable efficient comparison of historical imagery to modern imagery , which can augment basic change de tection and amplify other geospatial procedures for applications such as environment al assessments, cadastral information , and landscape changes . The great appeal of [Geo graphic Information Systems] GIS to historians is its ability to make historical maps commensurate with modern space and mapping conventions. Because of mistakes in projection even maps like the early USGS quads [DOQ], have to be georeferenced. We can geo reference or correct them so that they correspond to modern projections of the globe . 9 Historical maps, chart s , and images have a plethora of information imbedded in them, and how the information is extrapolated from the se data sources GIS is breathing new life into historical maps by freeing them from the static confines o f their 10 The acquisition of maps comes th rough different methods. S ome are released due to change in 8 (D'Este, 1983) 9 (White, 2010) 10 (Rumsey & Williams, 2002)

PAGE 20

20 classification (military intelligence imagery), and the end of regulations in gener al, which can release additional maps that would not be available before. A wealth of historic land cover/use maps, statistics have been produced and these are now more accessible due to the ending of copyrights and secrecy statuses, enthusiastic hobby co mmunities , and national cartographic institutes or cadastres that have a strategy towards data sharing with society. 11 Information extraction from historical geospatial data is often time s consuming , requiring many hours of processing. The subject matter al so influences the overall process because if access to the ancillary historical information is sparse or very limited , then one cannot always obtain the supporting/contextual maps, charts, or images required to process the information. Maps are offering valuable information related not only to spatial reference but also with respect to the time they are referred. In most cases, due to the scarcity of the geometric referencing and the lack of other relevant be inserted into routine GIS at least in a conventional sense . 12 Vertical and O blique P hotography Aerial photography during World War II was either conducted for reconnaissance operations or for map making, and most of the aerial images for this research were conducted by photographic reconnaissance units . M any images were taken within days of the invasio n for final planning and battlefield updates for the invading force s . The images used for this research were taken at both oblique and vertical angle s . T he oblique photograph is taken with the camera [purposefully] so held and aimed that its longitudinal axis forms an angle of less than 90 ° with the horizontal plane of the ground; the vertical photograph is taken with the camera pointed at the earth so tha t its longitudinal axis is 11 (Fuchs, Verburg, Clevers, & Herold, 2015) 12 (Balletti, 2006)

PAGE 21

21 perpendicular to the horizontal plane, or at least forms an angle as near 90 ° as possible. 13 Obliques were taken from either in side the aircraft pointing out or by hanging the camera out of the aircraft. The F 5 (P 38 Lightning variant) and P 51 Mustang ut i lized the inside configuration, the F 3 Havoc applied the overhang method, and the B 17 Flying Fortress used both the inside and overhang methods . Most of the obliques investigated appear to be taken inside the aircraft towards the side of the aircraft, as the propeller can be easily identified in the images. Figure 1 2 shows an aerial taken the day of the invasion with the propeller in the image. The oblique imagery could be taken at a lower elevation due to the angle at which the imagery was taken and did not require the airframe to be directly over the imaged area. This allowed many images to be taken at a height of 200 above ground level . The obliques are useful for determining height, observing large areas, or other c ircumstances that required a lower flight altitude to get the desired image . However, with smaller o verlap due to shorter camera to object distance in the oblique imagery, vertical images are simpler to align and build 3D models and orthomosaics. A vertica l aerial photograph is one made with the camera held or suspended in the airplane so that it points directly downward, and its longitudinal axis is therefore perpendicular to the surface of the earth. When the ground is flat and at the moment of taking suc h a photograph the airplane is level so that the surface of the film is parallel to the surface of the earth. 14 Vertical photographs were taken at a higher elevation, as they were used to cover larger area s. If vertical photograph s are captured with the sam e lens, then the only change in image scale is due to the height of the air plane above ground. If there are restrictions on 13 (US War Department Air Corps, 1938) 14 (US War Department Air Corps, 1938)

PAGE 22

22 flying height (such as flying higher to avoid enemy fire), scale can be changed by adjusting focal length with longer focal lengths leading to larger scale imagery . The Army Air Corps standard during WW II was a 60% forward overlap for each subsequent vertical image , and many of the vertical images used for this study met or exceeded that standard using the viewfinder in Figure 1 3 . Sid e overlap was not a standard established in the US Army Air Corps during WW II. Th e 60% forward overlap allows for not only mosaicking but building an orthomosaic and making maps via stereo compilation . The distance between these two lateral lines [in the viewfinder] varies according to the focal length of the lens of the camera with which the plate is designed to be used, the separation of the lines being such as to secure at any altitude a 60 percent overlap of each photograph over the preceding o ne in a strip of vertical aerial photographs. 15 The overlap seen in the images is pertinent to the present research and helps identify if the images can easily be aligned with each other and are good candidates for dense point cloud generation . The US Army Air Corps showed overlap was important , especially when flying at a higher altitude , to building maps for mapping or reconnaissance. Fiducial marks are reference points in large aerial film that define an image coordinate system which can be used for measu ring features within the photos and compensating for camera distortion . Cameras can have different fiducial mark locations, either in the middle of the sides of the image or in the corners . T he K 17 camera s used for the WWII reconnaissance investigated in this study have fiducials at the midpoint along each edge of the image. 15 (US War Department Air Corps, 1938)

PAGE 23

23 Fiducial marks in all aerial cameras are four in number and are either located in the four corners of the negative opening or in the center of each of the four sides. In either case, t he intent is to locate the marks so that lines joining opposite fiducial marks will intersect at the principal point of the negative. 16 Not all fiducial markings are the same and they are dependent upon different cameras, but all the K 17 images acquired us of four half arrows in the middle [ of the edges] in two different sizes. A pair of bigger half arrows indicate the flight direction. The other pair is intentionally smaller in order not to occupy too muc 17 Figure 1 4 provides an aerial image with fiducial marks. The fiducial marks not only identify the principal point and establish a predetermined point from which measurements and processing can be taken, but they allow for a determina tion of image measurement s cale (via calibrated distance between them). Thus, the physical coordinate units (e.g. m m ) of images on the photos can enable the use of physical units for focal length for subsequent processing . This allows for accurate initial approximation of focal length via the nominal values provided (e.g. f = Fairchild K 17 Camera The Fairchild K 17 camera was the primary camera used by the Army Air Corps for photo reconnaissance during WW II and it was the premier aerial photographic camera at the time. 17 camera manufactured by the Fairchild Camera and Instrument Corporation has been widely used for mapping in recent years [ 1944 1949 ] , 18 Figure 1 5 shows a K 17 16 (American Society of Photogrammetry, 1944) 17 (Redweik, Roque, Marqu es, Matildes, & Marques, 2009) 18 (United States Department of Commerce, Coast and Geodetic Survey, 1949)

PAGE 24

24 lens. There was a n R 17 used by the UK Royal Air Force that was the exact same camera as the K 17 but used a different nomenclature to identify a British photographic camera from a U.S. photographic camera. The camera took a 9 x 9 photograph , and could employ lenses. At the time , the onl y company es was Bausch and Lomb . T es w ere produced by East Lomb and Eastman Kodak. The images taken were from a film magazine that fit on top of th e camera and could be loaded and unloaded while in the air. This simplified the process of aerial photog raphy as it did not require the plane to land to reload the fil m. The usual magazine capacity wa s 250 exposures, but the manufacturer could supply a magazine with a capacity of approximately 500 exposures. Most of the film magazines found during the resear ch produced around 250 exposure s , but it is possible that 500 exposure film canister s were used for Normandy reconnaissance . Figure 1 6 shows a cross section of a K 17 camera and the location of film spools and vacuum plate. The camera utilize d a vacuum plate as the main mechanism to secure the film along the focal pla n e and remove all air between the two. This allowed for the film to be located at an exact and repeatable distance from the lens , enabling via the fiducial coordinates precise im age measurements to be made . After WWII, a modified K 17 camera renamed the was utilized by military and federal government agencies , includ ing the U S Coast and Geodetic Survey, USGS, and USDA. 17] camera is accurate

PAGE 25

25 19 Military entities The re are different military entities that impacted the coast of Normandy (France) , to incl ude naval vessels and aircraft units. The military vessels included many different warships, including aircraft carriers, battleships, destroyers, submarines , and amphibious warships . Most of the vessels were operated by the Navy, but t he amphibious warshi ps were Army vessels. Naval vessels The US Navy had three battleships that partic ipated in the Normandy landings; USS Arkansas (Eastern Omaha Beach), USS Nevada (Utah Be ach), and USS Texas (Western Omaha Beach). USS Texas attacked the target area of this research , and during the initial bombardment , which lasted approximately 34 minutes , the USS Texas launched 255 14 inch shells. There are no would cause, but the naval gunfire would travel from a northern directi on and cause the impact to have higher earthen dirt dispersion on the southern portion of a crater. The craters near th e target area can be attributed to either the naval gunfire rounds or aerial bomb ardment units . A ircraft units B omb ardment units conducte d aerial bombings over the beaches near Normandy; this include s the B 17 Flying Fortress aircraft , which w as later transitioned to an aerial re connaissance aircraft. A erial bombardment unit impacts would c reat e a 19 (United States Department of Commerce, Coast and Geodetic Survey, 1949)

PAGE 26

26 spherical shape d crater in the ground. This is a different shape from the directional crater caused by the naval gunfire rounds. The US Army Air Corps was the primary aerial reconnaissance group during the Normandy Invasion , although the British Royal Air Force contributed as well. Table 1 1 provid es a list of the combat squadrons in the US Army Air Corps that operated combat photo reconnaissance over Normandy from 1943 1945. The F 5 ( P 38 Lightning variant) , P 51 Mustang, and the F 3 Havoc were fighter and photo reconnaissance aircraft prior to the war, but the B 17 Flying Fortress was retrofitted during the war for photo reconnaissance. Figures 1 7 through 1 10 are images of the aircrafts. All the aircraft operated the K 17 camera for pre invasion , during i nvasion , and post invasion imagery . There was a Fairchild K 20 camera that which was in operation during the war, but it was not utilized for any combat missions during or within a year of the Normandy invasion. The K 17 camera cap tured imagery using a x format . The K 17 camera could be used for both vertical and oblique imagery, but the oblique imagery could only be acquired with len s sizes would hang over the aircraft too much .

PAGE 27

27 Table 1 1. Table of Combat Squadrons by aircraft flown during the combat missions flown over the beaches near Normandy 20 Aircraft F 5 ( P 38 Lighting variant) B 17 Flying Fortress 13th Photographic Reconnaissance 364th Bombardment 30th Photographic Reconnaissance 365th Bombardment 31st Photographic Reconnaissance 366th Bombardment 33d Photographic Reconnaissance 367th Bombardment 34th Photographic Reconnaissance 368th Bombardment 369th Bombardment 422d Bombardment 423d Bombardment Aircraft P 51 Mustang F 3 Havoc 33d Photographic Reconnaissance 155th Photographic Reconnaissance 111th Tactical Reconnaissance 20 (Department of the Air Force, 1982)

PAGE 28

28 21 Figure 1 1. The invasion beaches and the crossing routes across the English Channel ( By courtesy of Encyclopædia Britannica, Inc., copyright 2012; used with permission ) 21 (Encyclopædia Britannica, 2020)

PAGE 29

29 22 Figure 1 2. Aerial imagery day of invasion with propeller (Photo courtesy of the US Army Air Corps) 22 (Aerial Photograph, Can #DN5596, EX 0063, 1944)

PAGE 30

30 23 Figure 1 3. Viewfinder used by K 17 camera to ensure required overlap (Photo courtesy of Dirk Spennemann) 23 (Fairchild K 17A Aerial Camera, 2011)

PAGE 31

31 Figure 1 4. Aerial image with fiducial marks 24 (Photo courtesy of the US Army Air Corps) 24 (Aerial Photograph, Can #DN5851, EX 0138, 1944)

PAGE 32

32 Figure 1 5. K 17 camera 25 (Figure c ourtesy of the US Department of Commerce ) Figure 1 6. Cross section of K 17 camera 26 (Figure courtesy of US War Department) 25 (United States Department of Commerce, Coast and Geodetic Survey, 1949) 26 (War Department, 1944)

PAGE 33

33 Figure 1 7. Image of the P 38 Lightning 27 (Photo courtesy of the US Army Air Corps) 27 (Corps, Pilot Training Manual for the Lightning P 38, 1945)

PAGE 34

34 Figure 1 8. Image of the P 51 Mustang 28 (Photo courtesy of the US Army Air Corps) 28 (Corps, Pilot Training Manual for the Mustang P 51, 1945)

PAGE 35

35 Figure 1 9. Image of the F 3 Havoc 29 29 (Creative Commons, 2020)

PAGE 36

36 Figure 1 10. Image of the B 17 Flying Fortress 30 30 (Creative Commons, 2020)

PAGE 37

37 CHAPTER 2 METHODS The primary workflow steps for this project w ere 1) image retrieval, 2) initial processing and image conditioning, and 3) photogrammetric processing to develop geospatial products. The image retrieval process s indexing scheme for WWII photography, identifying the photos of interest, requesting transport of any film located at a different location/facility, visiting the facility, and scanning the imagery to produce digital files for use in processing . Figure 2 1 show a workflow of the steps for this project. The initial processing involve d file conversion, scaling, transforming, and clipping the imagery based on nominal physical dimension s . Image conditioning encompasses digital noise filtering and artifact removal. The photogrammetric component of processing was carried out using commercially available structure from motion (SfM ) photogrammetric software that provides automatic image matching, camera calibration, optimal image al ignment, dense 3D point cloud generation, and orthophoto production. Image R etrieval There are a limited number of WWII images of Normandy and the beaches available online, but without the parameters and multiples to overlap, the images a re unusable in th e context of this research . Thus, the aerial images were acq uired from NARA and NCAP with all the images captured f o r aerial reconnaissance purposes prior to D Day . A cquiring the images required very specific , time consuming methodology. Figure 2 2 shows a workflow of the NARA image retrieval steps. The steps began with searching the NARA website for the image bin, or record group (RG) , that contained the images taken by the government in the timeframe of the war, and in the area of the

PAGE 38

38 Normandy beaches. Th e initial search on the NARA web site identified that record group 373 ( RG 373 ) held the images for the timeframe around W W II and the Normandy I nvasion. RG 373 contained flight specific indexes ( overlays ), and they all f e ll under the larger bin for the Department of Defense (DOD), Defense Intelligence Agency (DIA), C entral I magery P rocessing and R eference D ivision. The search then required investigation to identify the grid squares that contained the Normandy beaches. Degree Square 49N000W (i.e. 49ºN La titude, 0º W Longitude) was identified first and showed many overlays that contained historical imagery of Omaha, Gold, Juno, and Sword beaches . Degree square 49N001W contained historical imagery of Utah and Omaha. Th ese degree square s fall in the imagery b in of overlay indexes for aerial photography of the DIA. I dentifying the overlays that may contain images of the Normandy beaches, required superimposing each overlay over an individual line map of the degree square. The image overlay is a representation o f the field of view from the camera at the instant the exposure was taken, and the overlays identify specific exposures numbers to assist in identifying a specific image. Figure 2 3 shows the 49N000W map, Figure 2 4 shows an image overlay example , and Figure 2 5 shows the overlay placed o ver the 49N 000W map. Initial Overlay I dentification The next step was recording all the pertinent meta data from the overlays that could assist with the analysis. The specific data fell into two categories : 1) photogram metry and 2) flight data. The photogrammetry data included focal length, fl ying height , and scale. The flight data include d o rganization, s ortie (aircraft), date, location, and quality of negatives. The flight data is helpful as it not only pro v id es us the ability to search for NARA negatives, but through identifying the unit it also helps identif y

PAGE 39

39 the camera that was likely used. This helps tie the imagery t o the K 17 or K 18 camera, and thus facilitates formatting , calibrat ing , and scal ing of the negatives for photogrammetric processing . Table 2 1 is an example of the format used for locating overlays and images to scan for processing. Overlay B each I dentification The next step in the initial image acquisition was to locate the images that the Army Air Corps took over the five beaches of Operation Neptune. Once the overlays were placed on the maps, then the image selection analysis began. One by one, each overlay was reviewed to identify which overlays contain ed the desired images. The Normandy invasion area spans two separate degree squares (49N000W, 49N001W) . T hey we re processed separately and then merged to determine possible overlaps in flights. The beach identification verified that the 49N000W degree square contained im ages of Omaha, Gold, Juno, and Sword beaches , and t he 49N001W deg ree square contained Utah and Omaha beaches. Scan of I dentified O verlays / I mages O nce identification of the images was completed , the next step was to create a digital scan of the film nega tive. This required multiple trips to the National Archives located in College Park, Maryland. The scanner utilized was a Microtek Scanmaker 9800XL Plus, and the scan settings used were designed for the reproduction of negative film gray scale at 1200 poin t s per inch (ppi) with the negative settings aligned to Kodak Supra film. 1200 ppi was chosen to maximize resolution given the limited time available at NARA since an increased amount of time would be required for higher resolution.

PAGE 40

40 A critical step before scanning was to check , by visual inspection, whether the images included the correct location , which they sometimes did not . After scanning, it was verified that the scanned image was properly generated without artifacts and with sufficient brightness/con trast . Prep rocessing The steps for preprocessing include 1) image conversion, 2) fiducial measurements, projective transformation , and artifact and noise removal. Quantum Geospatial Information System (QGIS) , software was utilized to complete a batch imag e conversion from a tagged image file format (TIFF) to bitmap (BMP). AutoCAD, an image processing software , was used for pixel coordinate identification. ImageJ , an open source image processing program utilizing Java scripts and plugins , was used for artif act and noise removal . The photogrammetric processing steps include 1) aligning photos, 2) optimization, 3) creating a dense point cloud, and 4) generating an orthomosaic. AgiSoft PhotoScan is photogrammetric software that processes and represents modern SfM software, and Cloud Compare is an open source 3D point cloud software. Image C onversion QGIS and P hotoshop w ere used in pre processing imagery for subsequent operations. QGIS was used for batch processing conversion of images from a TIFF to a n 8 bit BMP format, a file type requirement for t he resampling software used later, via the raster translation software . It is worth noting that the original scans were 8 bit images, so there was no lo ss in radiometric resolution. Once the image wa s an 8 bit BMP, then P hotoshop was used transpose the image by flipping either it horizontally or vertically or rotat ing the image 90º. This transposition was not required on all images ,

PAGE 41

41 because most images ha d the US Army Air Corps identification label located in the bottom left , indicating correct orientation . Figure 2 6 demonstrates the correct location of the US Army Air Corps identification label. Fiducial Measurement The next step was to identify the pixe l coordinates of the fiducials in AutoCad . The fiducial coordinates were manually located using the intersection o f the horizontal and vertical edges of the fiducial markers, which serve d as their calibration reference locations (i.e. the location from which the calibrated distance between fiducials are measured) . A locator/identifier separation protocol (Lisp) program , written by Dr. Bon DeWitt, was used to export the fiducial coordinates to a text file . The fiducial coordinates are essential for the following step. Projective Transformation Since the images were scanned, they were not aligned with each other relative to their fiducial coordinate system. Two computer programs developed and written by Dr. Bon DeWitt were used to calculate 2D projective transformations for flight (and therefore camera) specific images and then resample them so that they all had the same dimension and orientation with respect to their fiducials. This resulted in images that mimicked digital photographs taken from the same digital camera, allowing for camera calibration parameter solution in the photogrammetric processing steps. satellite im aging sensors, digital cameras, and scanned aerial photographs. In its rudimentary form, a digital image bears no relationship to a ground coordinate reference system. Rather, its coordinate basis consists of integer column and rows numbers which specify a location with a rectangular image array. 1 1 (Wolf, DeWitt, & Wilkinson, 2014)

PAGE 42

42 The transformation step utilized a 2D projective model to account for some of the distortions introduced by the scanner. The actual calibration parameters, including the calibrated fiducial coordinates, we re not known since calibrations reports were not available for the images . Instead , the assumption was made that the format was square. Based on inspection of calibration reports for other K 17 cameras, the format is likely very c l ose to square. For exampl e, an example calibration repor t showed corner to corner distance of 314.42 mm, and fiducial to fiducial distance of 222.3 1 mm. The fractional pixel coordinates of the fiducials were manually measured in he fiducial were chosen so that resampling of the imagery would result in square images with fiducial r eference points along the edges and with the intersection of opposite fiducials occurring at the center of the image. The transformation was also designe d so that the resulting image would have slightly higher resolution than the input image to minimize the loss of information from down sampling. After the projective transformation was complete , the transformation parameters we re input in to a rectification program, which produce d the synthesized, geometrically consistent images. T he resampling process used bicubic interpolation to maximize fidelity. B icubic resampling uses a cubic spline approximation and mirrors a sin c function to give the best resampling method . The other two methods are nearest neig hbor and bilinear interpolation: n earest neighbor chooses the digital number closest to the center of the grid square, and bilinear interpolation selects the four surrounding pixels and interpolates the pixel value from them. Bicubic interpolation is the most rigorous resampling technique of the three [methods] on the basis of signal processing theory. It achieves the

PAGE 43

43 smooth appearance without sacrificing as much high frequency (edge ) detail as is characteristics on the bilinear interpolation. 2 Artifact and Noise Removal ImageJ software was used for noise reduction and light band reduction. Some of the scanned images contained bands of bright light. Figure 2 6 shows an example of the light bands present in the imagery. These light bands , which were not produced by the scanning process but were present on the film itself and only on seven images , caused brightness differences that lead to the failure of the image matching process ( requi red for automatic SfM photogrammetric processing ) . The bandpass filter in ImageJ efficiently c orrects these lines, but care must be taken to ensure the filter does not remove key details f rom the image ry . The bandpass filter enables attenuat ion of specific spatial frequ encies within the image . Frequencies approximating t hose exhibited by the artifact we re selected for removal while minimally affecting the desirable spatial frequencies, such as the fine scale detail around areas of interest, such as small craters caused by damage. The sett ings utilized for the images were different i n the pre invasion and post invasion imagery. The software allows parameterization of the bandpass filter by identifying maximum and minimum dimensions of features, in pixel units, to be targeted and removed from the imagery. Similarly, there is an option to suppress vertical o r horizontal stripes, which was only utilized on the pre invasion imagery. The pre invasion imagery parameters were chosen to filter out systematic artifacts larger than 1200 pixels, and post invasion imagery was filtered using parameters to remove artifac ts larger than 1 000 pixels. 2 (Wolf, DeWitt, & Wilkinson, 2014)

PAGE 44

44 The median filter in ImageJ, was used to reduce speckle noise in the imagery. The median filter operates using a moving window with dimensions specified by the user, with the medial brightness values within the window assigned to the central pixel position . This allows for a risk reduction of high and low values affecting the image, but also can overly blur the image i f the window size is too high. The larger the window size , the more the attenuation of high frequency features a nd items of interest , like craters that may be only 5 10 pixels wide , will be attenuated . The median filter window size for the pre invasion imagery was 3 pixels and the post invasion imagery was 1.5 pixels. Figures 3 4 and 3 5 represent the pre and post f iltered images. After the bandpass and median filter s were applied, the images were imported into the photogrammetry software for processing . USGS C amera C alibration R eports During World War II, the US Army Air Corps required camera calibrations, but no documentation of these w as obtainable for the present study . USGS maintains K 17 camera calibration records starting in 1951 through the 1980s, and NARA maintains the US Army Air Corps camera calibrations, including any possible K 17 camera calibrations, prior to 1951. There were no identifiable K 17 records at NARA of any camera calibration reports. T he difference in principal point measurements from 1951 and the 1980s was very small , but the difference in fiducial coordinates was not recorded on the repo rts until 1978. compared from the 1950s and 60s, to the 1978/1980/1981 calibration reports, and similarly compared the principal points recorded. The calibration reports indicated that the principal points from 1951 19 81 were all within 0.0 5 mm , or 2.38 pixels, of each other .

PAGE 45

45 T herefore, the identified fiducial coordinates from the later reports could reasonably be applied to the earlier Normandy imagery for processing . Photogrammetric Processing Agisoft PhotoScan was chosen for photogrammetric processing of the imagery as it is one of the leading modern automatic SfM based software packages. I t i s also ubiquit ous , enabling the workflow developed here to be followed by a large number of practitione rs using it or similar software . The overall processing steps within PhotoScan include aligning the photos, self calibrating bundle adjustments (optimization), creating a dense point cloud, and generating an orthomosaic. Intermediate , but not required st ep s , include creating a mesh or tiled model. Image A lignment The first step was alignment of the images, which includes identif ication of conjugate features in overlapping photos, a nd solution of the relative position of camera stations . The US Army Air Corp s required forward overlap for vertical reconnaissance photography of 60% allowed for initial alignment of some of the images . Aligning the photos requires overlap . However, due to 100% of the during invasion photos being oblique images with little or no overlap , none of this imagery was usable in the photogrammetric portion of the workflow presented . The next processing step s included entry of the fiducial marks and coordinates , focal length, pixel size, and ground control points. T he fiduci al marks and fiducial coordinates were entered based on the average fiducial coordinates from the 1979/1980/1981 camera coordinates. F ocal length was based on the nominal values recorded on the image overlay s and converted from inches to millimeters . The p ixel size was calculated based on the following : the physical

PAGE 46

46 images ( 228.6 mm ) , t he image size wa s 10,580 x 10,580 pixels , and the image pixel size is divi ded by the pixel dimension . This calculation gives a size of 0. 0 2 2 mm per pixel. Equation 2 1 shows the calculation used. The ground c ontrol point coordinates were imported, and their locations were marked in the photos . The reference g round coordinate system must be established to allow for the correct optimization, which was WGS 1984 / UTM Zone 30N. Next, geographic position s of features were located that were visible on as many of the images as possible and established as control poin ts , including at intersections in the roads , intersections of fields, and near battle fortification s . All these points still exist to this day when viewed on current map ping me dia . Figure 2 7 depicts ground control points on Google Earth. The ground control point coordinates we re extract ed from both Google Earth Pro and DigitalGlobe . DigitalGlobe and Google Earth both use the WGS 84 for horizontal and vertical data and SRTM 2 data for height. SRTM 2 i s high resolution topographical data generated by t he space shuttle and supporting NASA satellites. Google states that it does not represent the most accurate geolocational data, and therefore is not the recommended method to use for control points. However, experimentation , by a JPL led team, was carried out with using G oogle E arth coordinates since they represent a readily available control coordinate source. If the accuracy is evaluated using all [checkpoints] CPs, regardless of their collection method and data provider, [Google Earth ] GE's imagery hor izontal positional accuracy over rural areas is 5.0 m RMSE, with mean horizontal error of 4.1 m and standard deviation (SD) of 2.9 m. If the accuracy is evaluated using all CPs, regardless of their collection method and data provider, [ DigitalGlobe ] DG's i magery horizontal positional ( 2 1 )

PAGE 47

47 accuracy over rural areas is 4.2 m RMSE, with mean horizontal error of 3.4 m and standard deviation (SD) of 2.3 m. 3 Due to their superior accuracy, DigitalGlobe coordinate values were ultimately used for the control points that were imported to georeference the aligned photos and assist via the optimization process . Optimization A USGS national UAS project office workflow , provided in Figure 2 8, was used to inform all steps in the photogrammetric processing . This workflow includ es parameters for semi automatic selection and removal of potentially spurious tie points. The parameters are reconstruction uncertainty , projection accuracy , tie point accuracy , and reprojection error . Deleting points based on reconstruction uncerta inty enhance the visual appearance and do not affect the optimization accuracy. Projection accuracy allows the filtering ou t of points that do not agree with the estimated location of neighboring points . The recommended t ightening (lowering) of the tie poi nt accuracy changes the accuracy of the tie points by a n order of magnitude (1 to .1) , thereby improving the adaptive camera modeling accuracy . Removing points with high reprojection error increases the accuracy when optimizing , since potentially spurious points are removed from observation . These combined steps reduce error s and noise which improve s the subsequent products such as the dense point cloud . During the bundle adjustments and simultaneous camera calibration , adaptive camera model fitting feature was used to determine camera calibration parameters that could best be resolved based on photo geometry and tie point distribution . 3 (Paredes Hernandez, Salinas Castillo, Guevara Cortina, & Martinez Becerra, 2013)

PAGE 48

48 Photogrammetric P roducts Next, utiliz ing a n ultra high quality setting and an aggres sive depth filter, a dense point cloud i s generated . From the point cloud data, a mesh is created . Mesh surface type and face count settings are dependent upon and proportional to the number of points in the dense cloud : the more points , the higher the fac e count. Since reference data is implemented in the processing, the heights are incorporated and a digital elevation model (DEM) is produc ed. The orthomosaic i s generated based on the optimized exterior ori entation and camera calibration parameters, and dense cloud data to depict an overall stitched 3 D image of the aerial photos with tilt and relief displacement removed. The orthomosaic can be considered a planimetric map on which horizontal features can be located and measured . Dense Point Cloud Analys is Cloud Compare wa s used for point cloud analysis and provide d segmentation and scalar field processing to allow for 3D measurements to be made within the area of interest . CloudCompare is a 3D point cloud (and triangular mesh) processing has been extended to a more generic point cloud processing software, including many advanced algorithms (registration, resampling, color/normal/scalar fields handling) 4 4 (CloudCompare)

PAGE 49

49 Figure 2 1. Overall Workflow for project Figure 2 2. NARA Workflow for project

PAGE 50

50 Figure 2 3 . 49N000W Map (Figure courtesy of NARA)

PAGE 51

51 Figure 2 4 . Overlay 11 (Figure courtesy of NARA)

PAGE 52

52 Figure 2 5 . Overlay 11 superimposed on map (Figure courtesy of NARA)

PAGE 53

53 5 Figure 2 6. Example of light lines in imagery (Photo courtesy of the US Army Air Corps) 5 (Aerial Photograph, Can #ON020666, EX 5110, 1944)

PAGE 54

54 6 Figure 2 7. Google Earth point locations 6 (Google Earth, 2019)

PAGE 55

55 Figure 2 8. USGS PhotoScan workflow (Figure courtesy of the USGS)

PAGE 56

56 Table 2 1 . Example of the format used for locating overlays and images to scan for processi ng Quad Overlay Film Identifier Organization Sortie Focal Length Altitude (feet) Scale (approx.) Date Time 000W 13 D6794, D6795 US 76P (US 7 27) 4M 1262 6" 28,000 1:56,000 30 Apr 44 001W 11 D8732 US 30 10PG 331 24" 31,000 1:15,500 9 May 44 1110PG 001W 35 D8910 US 31 500 24" 30,000 1:15,000 27 May 44 2005B 000W 16 D7193 US 7 27SQ 1738 6" 2,000 Oblique 6 Jun 44 1030 000W 18 D7957 US 80 6" 3,500 1:7000 6 Jun 44 1000 000W 24 D7237 US 7 1773 6" 6,000 Oblique 7 Jun 44 1800B 001W 68 D5875 US 7 GP 13SQ 3137 6" Oblique 27 Aug 44 1320 001W 71 E692 368 BS 6020 2 1 154.00 mm 20,000 1:40,000 18 Jun 45 1343 1545 001W 77 A6722 9 AF 17 IPDP 12" 2,000 Oblique 18 Jun 45 001W 79 E675 367 BS 5026 2 1 154.5 mm 20,000 1:40,000 19 Jun 45 0708 0954 000W 51 E750 423 BS 8193 2 1 154.6 mm 20,300 1:40,600 29 Jul 45 0747 0839 Quality Utah Omaha Gold Juno Sword NARA Can Barcode Priority Pictures Good Y Y Y (94)ON012160, (95)ON012159 (94)T153926534, (95)T153926533 55 5 Excellent Y DN5832 27 5 Excellent Y Y Y Y Y DN5598 1 61 Poor Y Y Y Y Y DN5604 2 15 Fair Y Y Y Y DN5851 4 39 Excellent Y DN5837 36 10 Good Y ON023069 T153940600 34 5 Excellent Y Y DN5829 45 22 Good Y Y ON022318 T153942861 46 8 Good Y Y Y Y Y ON023078 T153940609 51 15

PAGE 57

57 CHAPTER 3 ANALYSIS AND RESULTS A visual inspection of the orthomosaics shows obvious change in the pre invasion orthomosaic and the post invasion orthomosaic. Figures 3 1 and 3 2 show the pre invasion orthomosaics (taken 09 May 1944) , and F igure 3 3 shows the post invasion orthomosaic (taken 18 June 1945) . The pre invasion imagery is not georeferenced due to the excessive noise and inability to produce points in and near agricultural fields within the target areas , mostly due to whiteout . The main areas exhibiting noticeable changes are south and east of Point du Hoc before reaching Omaha B each and they can be seen in Figure s 3 1, 3 2, and 3 3. The areas just sou th of Point du Hoc ha ve many more impact rounds from aerial bombing and naval gunfire. The area east of Point du Hoc and west of Omaha B each was used as a vehicle staging and driving area , which is shown by the tracks just inside the beachhead vicinity. Overlays and I mages The 49N000W degree square contained 57 total overlays with only 1 6 overlays of possible images over the beaches. Next, taking those 16 overlays, it was projected how many images co vered the coastline/beaches. In the 16 overlays projecte d , there were 3 27 possible images with 273 (8 3 .5%) of them as oblique images and 54 (1 6 .5%) near vertical images. 49N001W deg ree square contained 103 total overlays with 2 6 overlays of possible images over the beaches. Next, tak ing those 6 5 overlays, it wa s estimat ed how many images co vered the coastline/beaches. In the 2 6 overlays projected , 2 06 possible images were identified with 113 (5 4 .9%) of them as oblique images and 9 3 (4 5 .1%) near vertical images. The 49N 001W degree square contained more than twice the number of overlays, but due to the smaller coastlines/beaches

PAGE 58

58 target area, it only presented 38.1% of the overall possible images. There were 64 initial images that had no film canisters on hand at NARA. T his could be attributed to either them never b eing delivered by DIA or the negative rolls being lost or destroyed. There are 29 total film canisters contained in the 4 1 searchable overlays: canister s ON023043 , DN5851, DN5460, DN5843 , DN5596, DN5829, DN5837, and ON022874 contained multiple overlays; there was one overlay (49N 000W, overlay 13) that was split into two different can ister s, and the remaining film canisters contained the other 19 overlays. Table 3 1 shows the number of overlays per NARA can. Two overlays (DN5604, ON059877) were not cataloged correctly at NARA and did not include Normandy imagery; therefore, only 41 searchable overlays were useable of the original 43. T he overlays were reprioritized by timeframe based on whether they were pre , during , or post invasion. This repr ioritization aligned the searchable 41 overlays to pre (9), during (15), and post invasion (17), and the two incorrectly cataloged overlays were one each in the pre and during categories. There were 533 (327 000W, 206 001W) possible images in the 41 ov erlays, with each overlay ranging from 3 83 possible images. Since the overlays were reprioritized by timeframe, the images were subsequently reprioritized. This reprioritization aligned the searchable 533 images to pre (70), during (242), and post inva sion (221). The pre invasion images totaled 87 images with a single overlay , number 27 (001W 27) , identifying 37 additional images that were initially not identifiable using the overlay, but deemed relevant after visual inspection. The other overlays were consistent in identifying image relevance based on their projections. The final during invasion images totaled 49 images with eight overlays . The other seven overlays contained no viable

PAGE 59

59 images and result ed in th e other 68 images being unusable. Three overlays produced comparable results to their projections for images, but the other five overlays only produced 20 50% of the projected images. The post invasion images totaled 107 images with four overlays, one overlay contained no viable images due to t he negative being too dark and resulting in overexposure in scanning, and this resulted in 50 images being unusable. Very few overlays produced comparable or better image numbers to their projections, but the largest overlays produced only 15% of the proje cted images. The post invasion o verlays had approximately 30 35 image duplicate s between the 000W and 001W degree squares. Prep rocessing During the pre processing portion of the research, the pre invasion and during invasion imagery w as found to be inadequa te f o r 3 dimensi onal change detection analysis. There was less than 10 % overlap in 94% (49 images) of the during invasion obliq ue imagery . The other 6% ( 3 images ) were mislabeled by NARA and unavailable for recovery. Bandpass Filter The b andpass filter was crucial for eliminating the bright white lines present in the film negative. The bandpass filter was tuned t o account for the 1157 and 1139 pixel width white lines in the pre invasion imagery . Since t he lines did not appear in the same locations along the same flight , they were not due to a scanning issue. ImageJ was used to effectively minimize the vertical light lines in the pre and post invasion imagery. The post invasion imagery had some small lines , 30 40 pixels wide , and the low spatial frequenc y filter was 1000 pixels to account for bright lines and patches . T he smoothing filter , objects smaller than the pixel number are attenuated, for both pre invasion and

PAGE 60

60 post invasion imagery was 0 pixels. Therefore, the smoothing filter was not applied as a ll pixels are greater than 0 , and a ny increase in the smoothing filter setting would degrade fine scale details, for example, small craters would be totally smooth and eliminated . Noise Removal The median filter pixel value determines the kernel size to neighboring pixels . The median filter replaces the central pixel with the median value of the pixels within the window . The pre invasion imagery processing benefitted from a median filter at 3 pixel s , attributable to the low imagery overlap and excessive i mage noise , allow ing for more noise to be attenuated without smoothing the craters. The low imagery overlap of the pre invasion benefited from the 3.0 pixel size median filter due to the limited number of eventual tie points available for matching and alig nment, also, the larger pixel size allows for more pixels to be accounted for and non s ensitiv ity to high or low values. The post invasion imagery allowed for a median filter at 1.5 pixel s to allow for effective photogrammetric processing , since it had h ig h er overlap and less noise . Figure s 3 4 A D and 3 5 A D show the pre invasion unfiltered and filter ed and post invasion unfiltered and filtered images. Photogrammetric Processing The first attempt at running through the photogrammetric processing steps was unsuccessfu l, since t he software could not generate enough tie points . The pre invasion images were very noisy , likely due to age. The noise in the imagery lead to erro r s in the image matching to generate tie points, which we re subsequently propagated to object spac e. Propagation of these errors wa s exacerbated by the relatively long focal length of the camera in the pre invasion photos .

PAGE 61

61 focal length has a much larger object space error area with a slight ray displacement . Figur e 3 6 shows the difference in the pre invasion imagery mm) and the post invasion mm) object space error areas . The dense cloud created from the pre invasion images had severe noise in the he ight comp onent and contained many artifacts that w ere not present in the during o r post invasion imagery. The noise and artifacts were the result of errors propagated from error in the image matching phases of both alignment and dense point cloud generation. Figure 3 7 A B depicts the difference in filtering with a dense point cloud with noise/artifacts (A) and a dense point cloud without (B) . Figure 3 7 B is a view of the dense point cloud facing the cliff, depicting the less noisy dense point clo ud with treetops and cliff height data to show the terrain change. The median/bandpass filters, along with despeckle convolution 3 x 3 matrix in ImageJ could not sufficiently alleviate the issues caused by the noisy film to generate useable geospatial prod ucts for the pre invasion imagery . Results and A nalysis Historical imagery can effectively be superimposed onto contemporary imagery. Using photogrammetric processing, t he post invasion Normandy analog historical images produced an orthomosaic that may be overlaid upon a current view of Google Earth. F igure s 3 8 and 3 9 show the ov erlaid orthomosaic with Figure 3 9 referencing the control/check points used. The pre invasion imagery was inadequate for successful process ing using the described processing step s due to excessive noise. However, t he post invasion imagery processing was successful , including geor eferencing vi a control points. The post invasion imagery , after photogrammetric processing , generated a dense point cloud

PAGE 62

62 using an ultra high quality se tting , a 3D model ultra high q uality setting , a DEM at 41.2 c enti m eters per pix el (cm/pix) ground sample distance (GSD) , and an orthomosaic with a GSD of 41.2 c m/pix . Table 3 2 shows the data point values during the photogrammetric processing. A 41.2 GSD means one pixel in the orthomosaic represents 41.2 cm on the ground. The calculated flight height by PhotoScan is 4270 m; therefore, the scale of the imagery is 1:28000. A 41.2 cm/pix GSD is the equivalent of a Phantom 4 drone flight at 1503 m. Control / C heck p oints Seven ground control points were selected from intersecting and easily identifiable locations, and after geolocation and processing , they created a PhotoScan calculated uncertainty of 0.18 m over the orthomosaic . There were 37 checkpoints selected across the orthomosaic to calculate the accuracy of the orthomosaic. Figure 3 9 shows the orthomosaic with the control and checkpoints displayed. Points six and 34 were not present in the orthomosai c; this can be attributed to the remo val of tie points near these checkpoints during the error reduction step , and that the checkpoints are near the edges of the images overlap. The 37 checkpoints were manually identified on the orthomosaic in PhotoScan and then they w ere compared to the chec kpoints visually identified on DigitalGlobe and Google Earth Pro . Table 3 1 shows the results of the analysis of the DigitalGlobe and Google Earth mapping engines. Both G oogle E arth and D igital G lobe use the Shuttle Radar Topography Mission (SRTM) for DEM d ata. The SRTM worldwide uncertaintie s are ± 10 m horizontal (longitude) , ± 10 m vertical (latitude) , and ± 8.7 m height (distance from geoid) , with a 90% accuracy. The Eurasia

PAGE 63

63 [specific] SRTM data shows uncertainty of ± 8.21 m horizontal , ± 8.23 m vertical, and ± 8.7 m height, with 90 % accuracy. 1 The checkpoint accuracy data is well within the SRTM accuracy errors for the horizontal, vertical, and height errors, but the Eurasia height RMSE is 3.7, and the calculated height RMSE was 3.85. This is t he only accuracy data point that did not exceed the SRTM error data. Table 3 3 shows the results of the checkpoint analysis of the DigitalGlobe and Google Earth points Crater / Cliff Size Naval gunfire and aerial bombers led an offensive as a precursor to and a part of Operation Neptune, and these craters are seen in many photos. They can be measured utilizing the products generated in this study . Cliff ed ge, size measurement and crater analysis , are measurable att ributes to validate the orthomosaic, and are completed using the dense point cloud created in PhotoScan by using C loud C ompare. Th e segmentation tool in CloudCompare allows users to remove excess point cloud data , and to focus on specific areas such as a cliff edge near the target area of this study . After the excess point cloud data is extract ed, the projection tools allows for the export at i o n of the coordinates to scalar fields i n side the segment ed portion , and the projection provide s height information . The m anipulation of the overall scalar field bounds allows representation of the edge of the beach (lowe r bound ) and top edge of the cliff (higher bound ) , enabling h eight calculation . 1 (Rodriguez, et al. , 2005)

PAGE 64

64 Figure 3 10 shows the cliff edge near the target area, after segmentat ion, scalar field projections, and height parameter manipulation. The control calculated hei ght value at the cliff edge is 6 0 m and the point cloud calculated height i s 60.06 m. A crater , most likely created by naval gunfire due to its proximity to a battle fortification and the shape of the crater, in the target area from this study is shown in Figure 3 1 1 . Figure 3 1 2 shows a current view of Google Earth with the crater shown at coordinates 30U 667655.21 m E, 5468459.75 m N . Figure 3 13 shows a n overview location of the crater. Figure 3 14 is an image of the crater from ground level taken Mar 2020. Figure 3 15 shows an updated image of the battlefield fortification in Figure 3 11. This crater was selected due to its size and the availability of p rojected depth data in the point cloud. After selecting the edge of the crater, t he crater volume was calculated using CloudCompare . The calculated volume was 46.28 m 3 . Using the volume of a partial sphere equation ( Equation 3 1), the software calculated v alue was validated with 46. 25 m 3 . Mass can be measured by taking the volume of the hole, and calculating the weight of the dirt displaced. D ensity for dirt is 76 lb / ft 3 . 2 Converting from ft 3 to m 3 , a lb of dirt weighs 2 683.56 lb s / m 3 . The volume calculated was 46.28 m 3 ; therefore, the dirt displaced was 46.28 m 3 x 2. 683.56 lb s / m 3 , and results in 124,195.16 lbs of dirt displaced from the crater. Figure 3 1 6 show s the volume calculations, and Figure 3 1 7 shows the crater inset with the crater volume grid. 2 (Engineering Toolbox, 2010) ( 3 1 )

PAGE 65

65 Table 3 1 . Number of overlays NARA can Number of overlays per NARA Can ON023043 4 DN5851, DN5460, DN5843 3 DN5596, DN5829, DN5837, ON022874 3 ON012160, ON012159, ON023078, DN5493, ON022318, ON023069, DN5844, DN5838, DN5603, DN5600, DN5595, DN5833, ON020666, DN5597, DN5850, DN5832, DN5846, DN5849, DN5825, DN5847, DN5592, DN5845, DN5598 1 ON059877, DN5604 0 Table 3 2. Photogrammetric processing data point values Data points Values Initial tie points 34,100 Tie points after error reduction 18,069 Dense point cloud points 152,331,027 3D model faces 30,442,213 DEM pixel size 12755 x 10591 Orthomosaic pixels size 12718 x 9560 Table 3 3 . Results of the check point analysis of DigitalGlobe and Google Earth Pro DigitalGlobe Google Earth Pro Horizontal mean error (m) 1.08 3.72 Standard deviation 4.26 4.88 RMSE (m) 4.21 3.85 Vertical mean error (m) 0.87 0.57 Standard deviation 2.95 4.05 RMSE (m) 2.92 4.82 Height mean error (m) 0.57 0.30 Standard deviation 3.90 4.25 RMSE (m) 3.85 4.00

PAGE 66

66 Figure 3 1. Pre invasion orthomosaic

PAGE 67

67 Figure 3 2. Pre invasion orthomosaic east of F igure 3 1

PAGE 68

68 Figure 3 3. Post invasion orthomosaic

PAGE 69

69 A B C D Figure 3 4 . Pre invasion filtered images A) 7114 3 B) 7115 4 C) 7116 5 D ) 7117 6 (Photos courtesy of the US Army Air Corps) 3 (Aerial Photograph, Can #DN5843, EX 7114, 1945) 4 (Aerial Photograph, Can #DN5843, EX 7115, 1945) 5 (Aerial Photograph, Can #DN5843, EX 7116, 1945) 6 (Aerial Photograph, Can #DN5843, EX 7117, 1945)

PAGE 70

70 A B C D Figure 3 5. Post invasion filtered images A) 7114 7 B) 7115 8 C) 7116 9 D) 7117 10 (Photos courtesy of the US Army Air Corps) 7 (Aerial Photograph, Can #DN5843, EX 7114, 1945) 8 (Aerial Photograph, Can #DN5843, EX 7115, 1945) 9 (Aerial Photograph, Can #DN5843, EX 7116, 1945) 10 (Aerial Photograph, Can #DN5843, EX 7117, 1945)

PAGE 71

71 Figure 3 focal length object space error

PAGE 72

72 A B Figure 3 7 . A) Dense point cloud with noise and artifacts , B) Dense point cloud after filtering with reduced noise and artifacts

PAGE 73

73 11 Figure 3 8 . Post invasion orthomosaic overlaid on current Normandy area imagery 12 Figure 3 9 . Post invasion orthomosaic overlaid on current Normandy area imagery with control and check points shown 11 (Google Earth, 2019) 12 (Google Earth, 2019)

PAGE 74

74 Figure 3 10 . Cloud Compare manipulated point cloud to show cliff height calculation 13 Figure 3 1 1 . Raw and processed image of the crater selected ( circled) and battle fortification in close proximity (triangle) (Photos courtesy of the US Army Air Corps) 13 (Aerial Photograph, Can #ON023043, EX 098, 1945)

PAGE 75

75 14 Figure 3 1 2 . Current Google Earth view of the crater selecte d 15 Figure 3 13. Overview of area with inset of crater 14 (Google Earth, 2019) 15 (Google Earth, 2019)

PAGE 76

76 Figure 3 14. Grenville Barnes. Normandy crater . March 03, 2020 . Longues sur Mer .

PAGE 77

77 Figure 3 15. Grenville Barnes. Normandy battle fortification . March 03, 2020. Longues sur Mer.

PAGE 78

78 Figure 3 1 6 . Volume calculation of crater

PAGE 79

79 Figure 3 1 7 . Volume calculated grid inset in crater

PAGE 80

80 CHAPTER 4 CONCLUSION Historical imagery , which can be challenging to obtain , can provide vital information to curren t times and assist with cadastral, military, and civili an studies/entities to better comprehend the environment al change . Historical imagery provides insight i n to the photogrammetric procedures at that time and show s how far photogrammetry has developed from WWII until now. Historical maps, chart s , and images have a plethora of information imbedded in them, and how th at information is extrapolated from the maps, charts, and images is important . Information extraction from historical geospatial data is often time consuming , requiring many hours of processing , bu t the results of the processing can produce information that was previously unknown. Whether the imagery is near vertical or oblique, the images can provide pertinen t information. Photogrammetric p rocessing of historical imagery create s georeferenced layer s before and after battles to compare to each other , which can play a paramount role to show how the war affected the area , including some naval gunfire craters that are still in existence today . The processing involves many steps including conversion, sc aling, transforming, clipping imagery, and conducting image conditioning. The photogrammetric component of processing uses commercially a vailable SfM photogrammetric software that provides automatic image matching, camera calibration, optimal image alignment, dense 3D point cloud generation, and orthophoto production. T his research establishes a proven workflow utilizing modern photogrammetry techniques and software to produce ph ot ogrammetric products from mid 1940s analog , single exposure film negatives, provided the appropriate level of overlap exists. Some images require more processing

PAGE 81

81 and transformations , as well as artifact and noise removal, but when processed they will help to collect information that may not be at tain ed other ways. Figures 3 1, 3 2, 3 3, 3 8 , and 3 9 reflect battlefield changes from pre and post invasion to include vehicle tracks and naval gunfire craters, and some of those changes (Figure 3 14) still exist today. The images show the landscape changes and how it was affected by the Normandy Invasion . This process can apply to po tentially denied or inaccessible areas that could be affected by conflict or natural disaster and show change s and measurements in t he desired area. The image retrieval performed in this study shows the challenge of obtaining the imagery needed to further aerial reconnaissance research . Different agencies holding similar records across the natio n makes it challenging to obtain desired records to conduct a complete and thorough analysis of the entire invasion areas. This workflow can serve as the framework to conduct further analysis with better imagery and product availability. This research could be expanded upon by researching more images from archives. If more records are obtained, then more research can be conducted. Similarly, better hardware could prevent washout and loss of quality of the film n egatives. However, as it is presented, t h is research accuracy assessment, and measurements of the cliffs and a crater prove the validity of the workflow and process, and these pr ocedures can be applied to battle damage analyses and assessment, as well as many other historical applications.

PAGE 82

82 APPENDIX A OPERATION NEPTUNE OPERATIONS ORDER Figure A 1. Operation Neptune operations order page 1 (Figure courtesy of the US Army Command and General Staff College)

PAGE 83

83 Figure A 2. Operation Neptune operations order page 2 (Figure courtesy of the US Army Command and General Staff College)

PAGE 84

84 Figure A 3. Operation Neptune operations order page 3 (Figure courtesy of the US Army Command and General Staff College)

PAGE 85

85 Figure A 4. Operation Neptune operati ons order page 4 (Figure courtesy of the US Army Command and General Staff College)

PAGE 86

86 Figure A 5. Operation Neptune operations order page 5 (Figure courtesy of the US Army Command and General Staff College)

PAGE 87

87 LIST OF REFERENCES Aerial Photograph, Can #DN5596, EX 0063. (1944, June 06). Aerial Flight Overlays for N 49 00 00 / W 00 00 00. Records of the Department of Defense. Defense Intelligence Agency. Central Imagery Processing and Reference Division, Record Group 3 73 (Aerial Photographs, compiled Sept 1943 to Oct 1945). National Archives at College Park, College Park, MD, USA. Aerial Photograph, Can #DN5843, EX 7114. (1945, June 18). Aerial Flight Overlays for N 49 00 00 / W 01 00 00. Records of the Department of Defense. Defense Intelligence Agency. Central Imagery Processing and Reference Division, Record Group 373(Aerial Photographs, compiled Sept 1943 to Oct 1945). National Archives at College Park, College Park, MD, USA. Aerial Photograph, Can #DN5843, EX 7115 . (1945, June 18). Aerial Flight Overlays for N 49 00 00 / W 01 00 00. Records of the Department of Defense. Defense Intelligence Agency. Central Imagery Processing and Reference Division, Record Group 373 (Aerial Photographs, compiled Sept 1943 to Oct 194 5). National Archives at College Park, College Park, MD, USA. Aerial Photograph, Can #DN5843, EX 7116. (1945, June 18). Aerial Flight Overlays for N 49 00 00 / W 01 00 00. Records of the Department of Defense. Defense Intelligence Agency. Central Imagery P rocessing and Reference Division, Record Group 373 (Aerial Photographs, compiled Sept 1943 to Oct 1945). National Archives at College Park, College Park, MD, USA. Aerial Photograph, Can #DN5843, EX 7117. (1945, June 18). Aerial Flight Overlays for N 49 00 00 / W 01 00 00. Records of the Department of Defense. Defense Intelligence Agency. Central Imagery Processing and Reference Division, Record Group 373 (Aerial Photographs, compiled Sept 1943 to Oct 1945). National Archives at College Park, College Park, M D, USA. Aerial Photograph, Can #DN5851, EX 0138. (1944, June 06). Aerial Flight Overlays for N 49 00 00 / W 01 00 00. Records of the Department of Defense. Defense Intelligence Agency. Central Imagery Processing and Reference Division, Record Group 373 (Ae rial Photographs, compiled Sept 1943 to Oct 1945). National Archives at College Park, College Park, MD, USA. Aerial Photograph, Can #ON020666, EX 5110. (1944, May 09). Aerial Flight Overlays for N 49 00 00 / W 01 00 00. National Archives at College Park, C ollege Park, md, USA. Aerial Photograph, Can #ON023043, EX 098. (1945, October 03). Aerial Flight Overlays for N 49 00 00 / W 00 00 00. National Archives at College Park, College Park, MD, USA.

PAGE 88

88 Albertz, J. (2002). Albrecht Meydenbauer Pioneer of Photogra mmetric Documentation of the Cultural Heritage. International Archives of Photogrammetry Remote Sensing and Spatial Information Sciences , 19 25. American Society of Photogrammetry. (1944). Manual of Photogrammetry. New York: Pitman Publishing Corporation. Balletti, C. (2006, Winter). Georeference in the analysis of the geometric content of early maps. e Perimetron , pp. 32 42. CloudCompare. (n.d.). Introduction. Retrieved October 15, 2018, from CloudCompare: http://cloudcompare.org/ Corps, H. A. (1945). Pi lot Training Manual for the Lightning P 38. Fort Wayne: Fort Wayne Printing Co. Corps, H. A. (1945). Pilot Training Manual for the Mustang P 51. Fort Wayne: Fort Wayne Printing. Creative Commons. (2020, March 03). B 17G "909". Retrieved from B 17 Flying Fo rtress Images: https://search.creativecommons.org/photos/bd45adb8 a888 4d8f 9d99 658f8fd6b5e5 Creative Commons. (2020, March 03). Douglas Havoc. Retrieved from Douglas Havoc Images: https://search.creativecommons.org/photos/b665b790 fdc8 48cb 95b7 9a6cea6b c7d8 Department of the Air Force. (1982). Combat Squadrons of the Air Force World War II. Headquarters USAF: Albert F. Simpson Historical Research Center and Office of Air Force History. D'Este, C. (1983). Decision in Normandy. New York: Diversion Books. E arth Resources Observation and Science (EROS) Center. (n.d.). USGS EROS Archive Aerial Photography Digital Orthophoto Quadrangle (DOQs). Retrieved 04 22, 2019, from USGS: https://www.usgs.gov/centers/eros/science/usgs eros archive aerial photography di gital orthophoto quadrangle doqs?qt science_center_objects=0#qt science_center_objects Encyclopædia Britannica, I. (2020, January 17). Normandy Invasion images and videos. Retrieved from Normandy Invasion, World War II: https://www.britannica.com/event/Nor mandy Invasion/images videos Engineering Toolbox. (2010). Dirt and Mud Densities. Retrieved February 19, 2020, from https://www.engineeringtoolbox.com/dirt mud densities d_1727.html

PAGE 89

89 Fairchild K 17A Aerial Camera. (2011, April 05). Retrieved February 13, 2020, from Fairchild K 17A Aerial Camera (sn# 44 84) nº 11: https://www.flickr.com/photos/heritagefutures/6361975269/in/pool camerawiki Fuchs, R., Verburg, P. H., Clevers, J. G., & Herold, M. (2015, May). The potential of old maps and encyclopaedias for re constructing historic European land cover/use change. Applied Geography , 59, pp. 43 55. Google Earth. (2019, July 02). Omaha Beach (30U 648401.72E 5472008.76N). Normandy, France. Greenspan, J. (2018, August 30). Landing at Normandy: The 5 Beaches of D Day . Retrieved April 18, 2019, from https://www.history.com/news/landing at normandy the 5 beaches of d day Hammond, W. M. (2003, October 03). Normandy: The Army Campaigns of World War II. Retrieved April 15, 2019, from Army History: https://history.army.mil/ brochures/normandy/nor pam.htm Min, S. (2020, February 19). War Thunder. Retrieved from [Vehicle Profile] DB 7 / Havoc (A 20): https://forum.warthunder.com/index.php?/topic/275356 vehicle profile db 7 havoc a 20/ National Archives and Records Administratio n. (n.d.). About the National Archives Museum. Retrieved 04 24, 2019, from National Archives Museum: https://museum.archives.gov/about National Collection of Aerial Photography. (n.d.). Our Work. Retrieved 04 01, 2019, from https://ncap.org.uk/about ncap/o ur work Nieuwint, J. (2020, February 19). 33 Beautiful Images of B 17 Flying Fortress In Flight. Retrieved from War History Online: https://www.warhistoryonline.com/military vehicle news/33 beautiful images of b 17 flying fortress in flight.html Paredes Hernandez, C. U., Salinas Castillo, W. E., Guevara Cortina, F., & Martinez Becerra, X. (2013). Horizontal positional accuracy of Google Earth's imagery over rual areas: a study case in Tamaulipas, Mexico. Boletim de Ciencias Geodesicas , 588 601. R edweik, P., Roque, D., Marques, A., Matildes, R., & Marques, F. (2009). Recovering Portugal Aerial Images Repository. ISPRS . Rodriguez, E., Morris, C. S., Belz, J. E., Chapin, E. C., Martin, J. M., Daffer, W., et al. (2005). An assessment of the SRTM topo graphic. Pasadena: Jet Propulsion Laboratory. Rumsey, D., & Williams, M. (2002). Past Time, Past Place: GSI for History. Redlands: ESRI Press.

PAGE 90

90 United States Department of Agriculture. (n.d.). Aerial Photography. Retrieved 10 15, 2018, from Farm Service Age ncy: https://www.fsa.usda.gov/programs and services/aerial photography/index United States Department of Commerce, Coast and Geodetic Survey. (1949). Topographic Manual Part II Photogrammetry. Washington D.C.: United States Goverment Printing Office. US Wa r Department Air Corps. (1938). Aerial Photography TM 2170 6. Washington D.C.: War Department. War Department. (1944). Aerial Photography. Washington D.C.: United States Government Printing Press. White, R. (2010, February 1). What is spatial history? Spat ial History Lab , pp. 1 6. Wolf, P., DeWitt, B., & Wilkinson, B. (2014). Elements of Photogrammetry with Applications in GIS. New York: McGraw Hill Education.

PAGE 91

91 BIOGRAPHICAL SKETCH N athaniel Balough completed his Master of S cience degree while serving as an active duty Army space officer. He completed his bachelor s degree in mathematics at North Greenville University in 2008. Nathaniel was the Treasurer of the UF Chapter for the Society of American Military Engine ers, an associate judge of the UF Traffic court, and a member of the Geomatics Student Association and the Natural Resources Diversity Initiative (NRDI).