Generating a 3D Point Cloud from UAV Images

Generating a 3D Point Cloud from UAV Images

Blue Marble Geographics



  2. Generating a 3D Point Cloud from UAV Images Kris Berglund,

    Vice-President, Sales
  3. • What is the LiDAR Module? • Automatic classification algorithms

    (ground, veg, buildings, powerlines) • Effective noise removal – filter for re-classification, delete entirely • 3D feature extraction, use cross-profile tool to define any object • QC tools for XYZ height adjustment, point cloud decimation • Global Mapper extension for advanced 3D point cloud processing
  4. • Create a high-density 3D point cloud from an array

    of overlapping images • Identify the same objects in multiple images from different perspectives • Generate precise orthorectified 3D image, create smoothed 3D mesh • Create optimized 3D point cloud from existing 3D mesh (less noise than original) • New point cloud thinning options – advanced decimation, ‘de-duping’ points
  5. • Active vs. Passive data collection techniques • Both have

    value, depending on project requirements • General comparison LiDAR vs. PhoDAR • Budget/cost and delivery time main decision making factors Light Detection & Ranging PhoDAR (photogrammetric PC) Requires flight planning UAS/UAV & ground pilot Laser scanner required RGB Camera (20Mpx) More expensive Less expensive Typically clean, sharp PC Usually processing required PC classifications exist No PC classifications exist
  6. • Over recent years, the availability and use of UAVs

    (Unmanned Aerial Vehicles) has rapidly expanded • Drones are now equipped with on-board GPS receivers, miniaturized cameras • 1-2 cm image resolution is relatively common • “Flying for Work” rules in U.S. are still evolving – FAA ‘commercially certified remote pilot’ – Visual line of sight – 400’ ceiling over target – Daytime only – 100 mph limit – Airport airspace restrictions UAS/UAVs and Drones
  7. • Landfill area volume reporting requirement (EPA, state agencies) •

    Regulators require periodic volumetric calculation of landfill area, project when capacity is reached • Landfill topography changes regularly • Drone flights using pre-set flight paths monitored by certified pilot • 200+ images acquired, georeferenced Project Challenge
  8. • Take a sequence of 2D photos using pre- defined

    flight plan • Import adjacent images from a range of different perspectives • ID same objects, RGB values • Recommended stereoscopic overlap: 75% - 85% • Triangulate using geotagged reference points • Use network of ground control points to tighten accuracy Structure from Motion – Pixels to Points
  9. • Acquire images, establish surveyed ground control points • Selection

    of correct images (removal of extra or erroneous images) • Create high-density 3D point cloud using Pixels to Points • Identify points for ground classification • Create Triangulated Grid – Use every point in the point cloud – Generates interlocking triangles to render an angular surface • Create Binned Grid – Apply smoothing based on sample area – Use one point at minimum elevation • Calculate volume based on area selection Process
  10. Generating a 3D Point Cloud from UAV Imagery Base layer

    showing target survey Pixels to Points (LiDAR Module) to generate 3D point clouds
  11. Add any files, folder of files, or loaded files Load

    established Ground Control Points Determine Output Option 3D point cloud Precise orthoimage 3D Mesh Optimized orthoimage
  12. Generating a 3D Point Cloud from UAV Imagery

  13. Generating a 3D Point Cloud from UAV Imagery

  14. Uses every point, creates interlocking triangles, generates angular surface

  15. Smooth grid based on local sample area, using one point

    with lowest elevation
  16. Incremental heights applied from polygon to create a plane to

    use as base elevation
  17. None
  18. None
  19. None
  20. None
  21. None
  22. • How Pixels to Points uses Photogrammetry & Triangulation •

    Generate point clouds w/EXIF geotags/georeferenced pixels • Use network of ground control points for tighter accuracy • Identify classified ground points • Grid DEMs from point cloud– Triangulate or Bin methods • 3D volumetric measurements, contour, area feature generation • 3D point cloud thinning for cleaner 3D mesh generation • Edit 3D meshes, split one mesh into many Summary – Pixels to Points workflow