UAS Photogrammetry Best Practices

Transcription

1 UAS Photogrammetry Best Practices Pennsylvania Society of Land Surveyors January 15, 2019 Bryan Baker Certified Mapping Scientist (UAS) Bryan Baker UAS Sales Manager Leica Geosystems Reality Capture Team Geospatial Instrument Industry 35 years. ASPRS Certified Mapping Scientist (CMS-UAS) Commercial Pilot Instrument Rated Remote Pilot Small UAS Factory Certified UAS Trainer FAA Safety Team (FAASTeam) Field Representative 2 1

2 Agenda ASPRS Digital Image Processing Standard Basic Photogrammetry Principles 3 ASPRS Positional Accuracy Standards for Digital Geospatial Data Positional_Accuracy_Standards.pdf 4 2

3 New ASPRS Positional Accuracy Standards for Digital Geospatial Data Replaces: ASPRS Accuracy Standards for Large-Scale Maps (1990) ASPRS Guidelines, Vertical Accuracy Reporting for Lidar Data (2004) Developed by: ASPRS Map Accuracy Standards Working Group, PAD, PDAD and LIDAR joint committee for map accuracy standard update In Final Approved Version REVISION 7, VERSION 1, Nov. 14, 2014 Approved and adopted by ASPRS during the board meeting on Monday Nov. 17, 2014 in Denver during ASPRS 2014 PECORA conference 5 New Standard for a New Era Motivation Behind the New Standard: Legacy map accuracy standards, such as the ASPRS 1990 standard and the NMAS of 1947, are outdated. (over 30 years since ASPRS1990 was written) Many of the data acquisition and mapping technologies that these standards were based on are no longer used. More recent advances in mapping technologies can now produce better quality and higher accuracy geospatial products and maps. Legacy map accuracy standards were designed to deal with plotted or drawn maps as the only medium to represent geospatial data. 6 3

4 New Standard for a New Era Within the past two decades (during the transition period between the hardcopy and softcopy mapping environments), most standard measures for relating GSD and map scale to the final mapping accuracy were inherited from photogrammetric practices using scanned film. New mapping processes and methodologies have become much more sophisticated with advances in technology and advances in our knowledge of mapping processes and mathematical modeling. Mapping accuracy can no longer be associated with the camera geometry and flying altitude alone (focal length, xp, yp, B/H ratio, etc.). 7 New Standard for a New Era New map accuracy is influenced by many factors such as: the quality of camera calibration parameters; quality and size of a Charged Coupled Device (CCD) used in the digital camera CCD array; amount of imagery overlap; quality of parallax determination or photo measurements; quality of the GPS signal; quality and density of ground controls; quality of the aerial triangulation solution; capability of the processing software to handle GPS drift and shift; capability of the processing software to handle camera self-calibration, the digital terrain model used for the production of orthoimagery. 8. 4

5 New Standard for a New Era These factors can vary widely from project to project, depending on the sensor used and specific methodology. For these reasons, existing accuracy measures based on map scale, film scale, GSD, c-factor and scanning resolution no longer apply to current geospatial mapping practices. Elevation products from the new technologies and active sensors such as lidar and IFSAR are not considered by the legacy mapping standards. New accuracy standards are needed to address elevation products derived from these technologies. 9 ASPRS Positional Accuracy Standards for Digital Geospatial Data Applicability: Defines specific accuracy classes and associated RMSE thresholds for digital orthoimagery, digital planimetric data, and digital elevation data Intended to be technology independent Limited to accuracy thresholds and testing methodologies for any mapping applications, and to meet immediate shortcomings in the outdated 1990 and 2004 standards Is not intended to cover classification accuracy of thematic maps Does not specify the best practices or methodologies needed to meet the accuracy thresholds Includes: Glossary, Symbols, examples, conversion to legacy standards 10 5

6 New Standard Highlights Positional Accuracy Thresholds which are independent of published GSD, map scale or contour interval digital orthoimagery digital elevation data Additional Accuracy Measures aerial triangulation accuracy, Ground controls accuracy, orthoimagery seam lines accuracy, lidar relative swath-to-swath accuracy, recommended minimum Nominal Pulse Density (NPD) horizontal accuracy of elevation data, delineation of low confidence areas for vertical data required number and spatial distribution of QA/QC check points based on project area. 11 Standard Deviation vs. RMSE Standard deviation is used to measure the spread of data around the mean. (Control Points) RMSE is used to measure distance between some values and prediction for those values. (Check Points) 12 6

7 New Standard Highlights It is All Metric! Unlimited Horizontal Accuracy Classes: Horizontal Accuracy Standards for Geospatial Data Horizontal Accuracy Class RMSE x and RMSE y RMSEr Horizontal Accuracy at 95% Confidence Level Orthoimagery Mosaic Seamline Mismatch X-cm X 1.41*X 2.45*X 2*X

8 15 Horizontal Accuracy Class RMSE x and RMSE y RMSE r Orthoimage Mosaic Seamline Maximum Mismatch Horizontal Accuracy at the 95% Confidence Level Common Horizontal Accuracy Classes according to the new standard [1]

9 Common Orthoimagery Pixel Sizes Associated Map Scale ASPRS 1990 Accuracy Class Associated Horizontal Accuracy According to Legacy ASPRS 1990 Standard RMSE x and RMSE y RMSE x and RMSE y in terms of pixels Examples on Horizontal Accuracy for Digital Orthoimagery interpreted from ASPRS 1990 Legacy Standard cm 1: cm 1: cm 1:200 5 cm 1: cm 1: cm 1:1, pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels pixels 17 Common Orthoimagery Pixel Sizes Recommended Horizontal Accuracy Class RMSE x and RMSE y Orthoimage RMSE x and RMSE y in terms of pixels Recommended use pixel Highest accuracy work Digital Orthoimagery Accuracy Examples for Current Large and Medium Format Metric Cameras 1.25 cm 2.5 cm 5 cm 7.5 cm 15 cm pixels Standard Mapping and GIS work pixels Visualization and less accurate work pixel Highest accuracy work pixels Standard Mapping and GIS work pixels Visualization and less accurate work pixel Highest accuracy work pixels Standard Mapping and GIS work pixels Visualization and less accurate work pixel Highest accuracy work pixels Standard Mapping and GIS work pixels Visualization and less accurate work pixel Highest accuracy work pixels Standard Mapping and GIS work pixels Visualization and less accurate work 18 9

10 Horizontal Accuracy/Quality Examples for High Accuracy Digital Planimetric Data ASPRS 2014 Equivalent to map scale in Horizontal Accuracy Class RMSE x and RMSE y RMSE r Horizontal Accuracy at the 95% Confidence Level Approximate GSD of Source Imagery ASPRS 1990 Class 1 ASPRS 1990 Class 2 Equivalent to map scale in NMAS to :25 1:12.5 1: to :50 1:25 1: to 2.5 1:100 1:50 1: to 5.0 1:200 1:100 1: to 7.5 1:300 1:150 1: to :400 1:200 1: to12.5 1:500 1:250 1: to :600 1:300 1: to :700 1:350 1: to :800 1:400 1: to :900 1:450 1: to :1000 1:500 1: to :1100 1:550 1: to :1200 1:600 1: New Standard Highlights Unlimited Vertical Accuracy Classes: Vertical Accuracy Standards for Digital Elevation Data Vertical Accuracy Class RMSE z Non- Vegetated Absolute Accuracy NVA at 95% Confidence Level VVA at 95 th Percentile Relative Accuracy (where applicable) Within- Swath Hard Surface Repeatability (Max Diff) Swath-to- Swath Non-Vegetated Terrain (RMSD z ) Swath-to- Swath Non-Vegetated Terrain (Max Diff) X-cm X 1.96*X 3.00*X 0.60*X 0.80*X 1.60*X 20 10

11 Vertical Accuracy/Quality Examples for Digital Elevation Data Vertical Accuracy Class RMSE z Non-Vegetated Absolute Accuracy NVA at 95% Confidence Level VVA at 95 th Percentile Within-Swath Hard Surface Repeatability (Max Diff) Relative Accuracy (where applicable) Swath-to-Swath Non-Veg Terrain (RMSD z ) Swath-to-Swath Non-Veg Terrain (Max Diff) 1-cm cm cm cm cm cm cm cm cm cm Vertical accuracy of the new ASPRS 2014 standard compared with legacy standards Vertical Accuracy Class RMSE z Non-Vegetated Equivalent Class 1 contour interval per ASPRS 1990 Equivalent Class 2 contour interval per ASPRS 1990 Equivalent contour interval per NMAS 1-cm cm cm cm cm cm cm cm cm cm

12 Examples on Vertical Accuracy and Recommended Lidar Point Density for Digital Elevation Data according to the new ASPRS 2014 standard Vertical Accuracy Class RMSE z Non-Vegetated Absolute Accuracy NVA at 95% Confidence Level Recommended Minimum NPD (pts/m 2 ) Recommended Maximum NPS 7 (m) 1-cm cm cm cm cm cm cm cm cm cm Horizontal accuracy requirements for elevation data Photogrammetric elevation data:, the horizontal accuracy equates to the horizontal accuracy class that would apply to planimetric data or digital orthoimagery produced from the same source imagery, using the same aerial triangulation/ins solution. Lidar elevation data: use the following formula: = + tan( )

13 Accuracy requirements for aerial triangulation and INS-based sensor orientation of digital imagery Accuracy of aerial triangulation designed for digital planimetric data (orthoimagery and/or digital planimetric map) only: RMSE x(at) or RMSE y(at) = ½ * RMSE x(map) or RMSE y(map) RMSE z(at) = RMSE x(map) or RMSE y(map) of orthoimagery Accuracy of aerial triangulation designed for elevation data, or planimetric data (orthoimagery and/or digital planimetric map) and elevation data production: RMSE x(at), RMSE y(at) or RMSE z(at) = ½ * RMSE x(map), RMSE y(map) or RMSE z(dem) 25 Accuracy requirements for ground control used for aerial triangulation Accuracy of ground controls designed for planimetric data (orthoimagery and/or digital planimetric map)production only: RMSE x or RMSE y = ¼ * RMSE x(map) or RMSE y(map), RMSE z = ½ * RMSE x(map) or RMSE y(map) Accuracy of ground controls designed for elevation data, or planimetric data and elevation data production: RMSE x, RMSE y or RMSE z = ¼ * RMSE x(map), RMSE y(map) or RMSE z(dem) 26 13

14 Examples on Aerial Triangulation and Ground Control Accuracy Product Accuracy (RMSE x, RMSE y ) Aerial Triangulation and Ground Control Accuracy Requirements, Orthoimagery and/or Planimetric Data Only RMSE x and RMSE y A/T Accuracy RMSE z Ground Control Accuracy RMSE x and RMSE y RMSE z Product Accuracy (RMSE x, RMSE y, or RMSE z ) RMSE x and RMSE y A/T Accuracy RMSE z Ground Control Accuracy RMSE x and RMSE y RMSE z Aerial Triangulation and Ground Control Accuracy Requirements, Orthoimagery and/or Planimetric Data and Elevation Data 27 Reporting Horizontal Accuracy This data set was tested to meet ASPRS Positional Accuracy Standards for Digital Geospatial Data (2014) for a RMSE x / RMSE y Horizontal Accuracy Class. Actual positional accuracy was found to be RMSE x = and RMSE y = cm which equates to +/- at 95% confidence level. This data set was produced to meet ASPRS Positional Accuracy Standards for Digital Geospatial Data (2014) for a RMSE x / RMSE y Horizontal Accuracy Class which equates to +/- cm at a 95% confidence level

15 Reporting Vertical Accuracy This data set was tested to meet ASPRS Positional Accuracy Standards for Digital Geospatial Data (2014) for a RMSE z Vertical Accuracy Class. Actual NVA accuracy was found to be RMSE z = cm, equating to +/- at 95% confidence level. Actual VVA accuracy was found to be +/- cm at the 95% percentile. This data set was produced to meet ASPRS Positional Accuracy Standards for Digital Geospatial Data (2014) for a cm RMSE z Vertical Accuracy Class equating to NVA =+/- cm at 95% confidence level and VVA =+/- cm at the 95% percentile 29 Recommended Number of Check Points Based on Area Project Area (Square Kilometers) Horizontal Accuracy Testing of Orthoimagery and Planimetrics Total Number of Static 2D/3D Check Points (clearly-defined points) Vertical and Horizontal Accuracy Testing of Elevation Data sets Number of Static 3D Check Points in NVA Number of Static 3D Check Points in VVA Total Number of Static 3D Check Points

16 Why ASPRS Mapping Standards To bring Survey Grade data reliability ASPRS positional accuracy standards used for reporting Using classical aerial photogrammetry checks Infinity Point Cloud Agisoft Point Cloud 31 Surface Flatness Analysis 32 16

17 Infinity Point Cloud Agisoft Point Cloud

18 35 Example on Applying the New Standard 36 No RMSEr 18

19 Example on Applying the New Standard 37 10cm RMSEr Example on Applying the New Standard 38 3cm RMSEr 19

20 Basic Photogrammetry Principles 39 Photogrammetry Definition Photogrammetry is science of taking reliable measurements from photography

21 Advantages of UAS Photogrammetry Photos provide a permanent record of what existed at the instant the pictures were taken. Photos are easily interpreted without the need for special training. Photos can be used for multiple purposes (i.e. inspection, measurement, dense point clouds, and ortho-rectified images). Terrain models and ortho-rectified images can be created relatively quickly and generally at a lower cost than ground surveying methods. UAS photogrammetry can be used to measure areas that would be difficult, dangerous, or even impossible to access from the ground. 41 Disadvantages of UAS Photogrammetry Photo acquisition limited by weather, time of day, and lighting conditions. Limited foliage penetration. You need to have the same location imaged in at least two photos in order to create a 3D Point. Advance permission required to commercially fly UAS in certain air space. Some locations completely off limits. Additional safety related restrictions like not flying UAS over non-participants or moving vehicles and line-of-sight limitations

22 Vertical vs. Oblique Photos Vertical Photograph A photo taken with the camera axis vertical or nearly so. Oblique Photo A photo taken with the camera axis intentionally tilted between vertical and horizontal. High Oblique Photo An oblique photo where the horizon is visible in the photograph. 43 Photogrammetry basics Pinhole Camera Model Image space Image plane For planning calculations, we can treat the camera lens system as a simple pinhole camera. Object space Reflective Surface Light emanating from the sun strikes the surface and reflected. Part of the reflected light enters the camera, passes through the focal center of the camera, and strikes the image plane

23 Photogrammetry Basics Photo Scale f By similar triangles d = = h H f : focal length H : Flying height d : sensor width/height D : footprint width/height D 45 Photogrammetry Basics Ground Sample Distance (GSD) p f H The Ground Sample Distance (GSD) is the dimension of the image pixel projected onto the ground. The same equation can be applied to calculate the GSD = = h GSD f : focal length H : Flying height p : pixel size GSD : Ground Sample Distance 46 23

24 Photogrammetry Basics Footprint size and GSD example = = 6 Camera: Sony alpha 7R II Focal Length: 28 mm Sensor size: 6000 x 4000 pixels Sensor dimensions: 35.8 x 23.9 mm Flying Height: 100 m h = h h = = = 21.4 = = 85.4 For your camera/lens combination, it s good to memorize these ratios so that you can quickly estimate the footprint size and GSD in the field for a given flying height. 47 Photogrammetry Basics Image Block Base Distance Interline Distance Flight Line Exposure Station 48 24

25 Photogrammetry Basics Image block In order to successfully orient a block of overlapping images so that a point cloud or ortho-rectified mosaic can be produced, the images must overlap by certain minimum amounts. Along the direction of flight, the images must overlap by a minimum of 60%. Between flight lines, the images must overlap by a minimum of 30%. For an added margin of safety to account for small terrain variations and imprecise camera positioning, minimum forward and side overlaps should be more like 70% and 40% respectively. Planned forward and side overlaps should not exceed 85% to 90%. 49 Photogrammetry Basics Base and Interline Distance Calculation B We can view this diagram as either two consecutive photos along a flight line or as looking down the ends of two adjoining flight lines. = 1 % = 1 % B %OV D 50 25

26 Photogrammetry Basics Base and Interline Distance Calculation Example Camera: Sony alpha 7R II Focal Length: 28 mm Sensor size: 6000 x 4000 pixels Sensor dimensions: 35.8 x 23.9 mm Flying Height: 100 m Forward overlap: 70% Side overlap: 40% = 100 = % = % = Mission Planning Image Overlap and Terrain Height Variation Decreased Image Overlap Terrain Variation Launch Elevation Flying over terrain that s rising up towards the copter causes a decrease in the planned image overlap to the point where downstream processing can be negatively affected. This overlap issue with upward terrain variations can be compensated for by increasing the planned percent forward and side overlap and/or increasing the flying altitude

27 Mission Planning Image Overlap and Terrain Height Variation Example Problem: Assuming a constant interline distance, what upward change in terrain height will cause the percent side overlap to decrease from 40% to the minimum 30%? H ΔH % = ( ) (1 30%) (1 40%) (1 30%) = Oblique Photo Geometry d/2 f α f: Camera focal length. d: Image frame dimension. α: Field of view angle. β: Depression Angle. H: Camera height. tan = 2 H β 1 = tan 2 = tan + L1 L

28 Oblique Photo Example = tan = 23 Camera: Sony alpha 7R II Focal Length: 28 mm Sensor size: 6000 x 4000 pixels Sensor dimensions: 35.8 x 23.9 mm Flying Height: 100 m Depression Angle (deg.) L1 L Relative vs. Network Accuracy Relative (aka Local) accuracy refers to the accuracy of distance measurements. A volumetric analysis can be performed as long as no external data needs to be incorporated such as a model of the underlying surface beneath a stock pile. Network (aka Absolute) accuracy refers to knowing the location of a point relative to an external reference system/datum like state plane or possibly a local reference system. Good network accuracy is required for analyzing changes over time and more generally when measurements must be integrated with existing data. Sometimes trying determine the coordinate system and datum of the data to be tied into is easier said than done 56 28

29 Ground Control Distribution Bounding the Project Area Project area bound by control. Project area not bound by control. For larger areas, be sure to scatter additional control points throughout the interior of the project to reduce warping of the point cloud. 57 Control Point Distribution Step 1 Bound the project area Place control points just to the outside of the project area at the inside and outside corners of the project. These control points should bound your project area

30 Control Point Distribution Step 2 Fill In Additional Control Place additional control points throughout the interior of the project making sure that the areas of influence of the control (yellow circles) overlap. Depending on the accuracy requirements of the project, whether GNSS photo positioning is used, its quality, and the quality of the camera sensor, the radius of the areas of influence may range from approximately 150 to 500 feet or more. At a bare minimum, there should be at least five control points well distributed throughout your project. 59 Ground Control Layout for Corridors Step 1 Set control on either side of corridor towards the edge of the photos at the ends of the flight lines and at major bends. Do not just place control down the centerline! The control must be placed so that so that it will fall in at least two photos

31 Ground Control Layout for Corridors Step 2 Fill in Additional Control Place additional control along the length of the project corridor alternating from side to side making sure that the areas of influence of the control (yellow circles) overlap. The control should be placed so that it falls towards the edge of the photo. Depending on the accuracy requirements of the project, whether GNSS photo positioning is used, its quality, and the quality of the camera sensor, the radius of influence may range from 150 to 500 feet. 61 Corridor Flight Planning Alternative Mission Plan Instead of flying just down the centerline of the corridor, consider flying along each side it such that the photos overlap side to side down the corridor. This keeps the aircraft from flying over traffic and at the same time improves the adjustment results because the two flights help support each other eliminating any side to side tilt that can happen along the flight line when only a single flight line is flown. The photos should have a side overlap of anywhere from 60 to 85%

32 Ground Control Target Design Target Size: Minimum of 5 to 10 times the nominal image GSD. Avoid targets that are so large as to be difficult to measure precisely on the photos. Color: Avoid unnatural, neon colors as these colors tend to bloom in the resulting images. It s best to stick to black and white targets for highest contrast. Recommended Target: 9 or 12 inch white or light gray ceramic tile with an X drawn on it and the opposing quadrants painted in black. They re durable, reusable, heavy enough not to get blown around, and no matter the flying height the center can easily be measured. 63 Natural Ground Control Targets Pro s: Positions can be recorded after the mission photos have been collected. Can be used to supplement targeted ground control or as check points. Don t leave any lasting marks like a painted target. Con s: Generally less accurate than targeted control. Can be more difficult to accurately locate. Natural ground control targets need to be Uniquely identifiable. Precisely locatable. Lie flat on the ground. Immovable. Visible from the air. Tip: In the field, take a photo of the control location to aid in identifying it s position when back in the office

33 Check Points Check points provide an independent check on the accuracy of the photo adjustment. They are typically collected at the same time as the control using similar criteria for placement, but they are not used as part of the adjustment process. Minimum of five required for a statistically significant sampling. Should be located half way between the control points. Do not place them right beside the control points. If only the vertical accuracy of the adjustment is to be evaluated, they do not necessarily need to be targeted but still located on a hard surface. 65 Camera Calibration Camera calibration is the process of precisely determining the internal characteristics of a camera system that can impact metric accuracy. These include the camera s focal length, optical center of the camera s sensor array, and geometric distortion of the lens system. These parameters can usually be automatically calculated by the orientation software. Caution should be exercised as these parameters, especially focal length, can be highly correlated with external factors. Without significant differences in the elevation of the ground control points within the project, the computed focal length may not be correct which can cause vertical scaling errors. Whenever possible, vary the height of the control points throughout the project area

34 Mission Planning Sun location and time of day The pilot should always position themselves with respect to the project area so that at no time during the flight does the copter get lost in the glare of the sun. A sun angle of between 30 and 60 degrees above the horizon is optimal. Sun angles lower than 30 degrees create long, deep shadows. Sun angles greater that 60 degrees create texturally flat images. 67 Mission Planning Wind Direction It s recommended that missions be planned such that the wind helps to bring the copter back towards its landing location on its last/return legs. At the end of a mission, the battery voltage is low and allowing the wind to help bring the copter back, extending the battery life, and increasing the odds of a safe return. Launch/recovery location Wind Direction Wind assisting on return legs

35 Mission Planning Using Every Leg Launch/recovery location Every leg of the mission should be utilized for collecting data. Instead of a wasted leg Launch/recovery location Increase the side overlap between flight lines and utilize all four legs. 69 Photography Basics Exposure Aperture Depth of Field A photo s exposure determines how light or dark an image will appear when captured. Exposure is dependent on just three camera settings aperture, shutter speed, and ISO. EXPOSURE ISO Image Noise Shutter Speed Motion Blur 70 35

36 Photography Basics Camera Settings Aperture: Controls the area over which light can enter the camera. Shutter Speed: Controls the amount of time light can enter the camera. ISO: Controls the camera sensors sensitivity. 71 Photography Basics Aperture Shallower Depth of Field Deeper The larger the f-number, the smaller the hole, and the deeper the depth of field. At the flying heights area based missions are typically flown, the camera is focused at infinity. The hyper focal distance will typically be much less than the flying height at all apertures so aperture is not a primary concern

37 Hyperfocal Distance H : Hyperfocal distance f : Lens focal length N : f-number c : Circle of confusion limit (0.03mm typically used). Hyper Focal Distance [m] The hyper focal distance is the distance beyond which all objects are acceptably sharp for a lens focused at infinity. Focal Length [mm] f-number Depth of Focus + s : subject distance. H : Hyperfocal distance. D N : Near focus distance. D F : Far focus distance. Example: An inspection flight is to be conducted with the camera focused at the planned image-object distance of 5 meters. The camera focal length is 50mm and the aperture is set to f/8 (Hyperfocal distance: 10.4 m) = = 9.6 Given the above, objects from 3.4 m to 9.6 m will be in acceptable focus

38 Photography Basics ISO Sweet Spot The quality of camera sensors vary, but typically ISO should be held to 800 or less. 75 Photography Basics Shutter Speed The faster the shutter speed, the less image blur due to motion

39 Photography Basics Forward Motion Blur Forward motion blur is caused by the aircraft s movement while the camera shutter is open. This effect can be reduced by increasing the shutter speed. h 2 For area mapping, the camera shutter speed should typically be 1/800s or faster. 77 Photography Basics Forward Motion Blur Example = = = = 3.2 Camera: Sony alpha 7R II Pixel Size: 6 microns Focal Length: 28 mm Flying Height: 70 m Shutter Speed: 1/500 s Aircraft Speed: 4 m/s Forward motion blur in this example would be just over half a pixel which would likely be visible in the image

40 Photography Basics Lens Focal Length Longer focal length lenses: Great for inspection flights because they allow the pilot to fly farther away from an object. Require more precise flying and good vibration dampening due to the narrower field of view. Shorter focal length lenses: Great for area mapping because they allow a larger area to be covered in a single frame. Too short of focal length can create excessive fisheye distortion. 79 Photography Basics Prime vs. Variable Focal Length Lenses Prime Better image quality. Lower light (faster). Simpler to use and set up. Lower cost. Smaller size and less weight. Variable Zoom More versatile

41 Sony Camera Comparison Camera Alpha 6000 Sony A7R Sony A7R II Image Sensor Width [mm] Image Sensor Height [mm] Pixel Size [um] Max Exposure Rate [frames/sec.] Sensor Width [pixels] Sensor Height [pixels] Pixels per square meter Processing Desnity.5cm GSD 1cm GSD 2cm GSD 3cm GSD 5cm GSD Full (High) 40,000 10,000 2,500 1, Half (Medium) 10,000 2, Quarter (Low) 2, Eighth (Lowest) Per square ft = PSM /

42 Semi-Global Matching (SGM) Introduced in 2005 by Heiko Hirchmueller from the DLR. A method of stereo matching on a pair of images using known intrinsic and extrinsic orientation of the cameras. A very robust algorithm used today in a wide range of computer vision applications, including navigation of autonomous vehicles. In addition to encoding the point cloud with RGB information, it can also be encoded with NDVI and NIR as well. Can generate very dense point clouds in which a 3D point can be generated from each pixel in the input images. 83 SGM Considerations Image Exposure SGM depends on the information contained in the imagery to perform its matching. Over and under exposed imagery creates noise in the resulting point cloud. Camera System Low quality lens systems can have excessive distortion which can carry through to the resulting point cloud. Lighting and Time of Day Subject A sun angle of between 30 and 60 degrees above the horizon is optimal. Sun angles lower than 30 degrees create long, deep shadows. Sun angles greater that 60 degrees create texturally flat images. Subjects that have little image texture or subjects with repeating patterns like farm field furrows can cause the image matching to generate excessive noise and mismatches

43 SGM Workflow Image Capture Care required to capture sufficiently overlapping Imagery to cover the project area. SGM Processing Well distributed ground control and/or GNSS positioning of the copter required to orient the imagery to the project coordinate system. Image Alignment Matching results improved with proper focus, exposure, and a shutter speed sufficient to minimize any image blur due to image motion. 85 Semi-Global Matching Increased point noise due to lack of image texture SGM Generated: 4cm spread Leica P50 Scanner: 0.4cm spread SGM Generated: 20cm spread 86 43

44 Semi-Global Matching Increased point noise due to poor exposure 87 Questions?? THANK YOU!! Bryan Baker Leica Geosystems Inc