UAV-based Structure from Motion for Sinkhole Characterization

UAV-based SfM

Why use SfM for mapping sinkhole characteristics?

What is SfM?

  • 3D from overlapping 2D images
  • Camera poses + sparse cloud → MVS dense cloud → DEM/orthomosaic

Point Cloud

Flight & Data

  • ≥75–85% overlap; nadir + oblique
  • GCPs or RTK/PPK
  • Consistent exposure; avoid harsh shadows

flight plan Camera Positions

Products

  • DEM/DSM, orthomosaic, point cloud, mesh
  • Derivatives: slope/aspect, contours, DoD (change)

3D Model DEM

Workflow Highlights

  1. Plan & fly
  2. Align images (tie points, bundle adjustment)
  3. Densify (MVS)
  4. Rasterize (DEM/DSM, ortho)
  5. QC (RMSE vs checkpoints)

For Sinkholes

  • Delineate rims/depressions
  • Metrics: depth, volume (cut/fill), slope, curvature
  • Integrate with lineaments & karst layers

Conclusion

  • We want characterize the karst landscapes of Arizona
  • We want to identify all the sinkholes and characterize theme
  • SfM is a fast and accurate way to survey sinkholes
  • Much more work is done on the backend to measure and categorize sinkholes after point cloud aquisition
  • There are many commercial and open source tools that can help with SfM