Dune migration and volume change from airborne LiDAR, terrestrial LiDAR and Structure from Motion-Multi View Stereo

摘要: Abstract Sand dunes are commonly regarded as a challenge to traditional photogrammetry due their homogeneous texture and spectral response. In this work we present an evaluation of Structure from Motion-Multi View Stereo (SfM-MVS) obtain high-resolution elevation data coastal sand based on images acquired by Remotely Piloted Aircraft (RPA). A Digital Elevation Model (DEM) dunefield in Southern Brazil was generated 810 photos captured RPA at 100 m above the takeoff point February 2019. Image matching successful all areas survey presence superficial features (footprints sandboard tracks) visibility sedimentary stratification, highlighted heavy minerals. Altimetric accuracy SfM-MVS DEM validated comparison with Terrestrial LiDAR (TLS) collected during same fieldwork campaign flights. The then compared Airborne (ALS) October 2010. While TLS DEMs very similar, without any major difference or reconstruction topographic features, presents small scale surface roughness not visible DEM. Feature Preserving Smoothing (FPD) algorithm applied good results terms smoothing, but significant changes descriptive statistics error metrics, RMSE 0.08 m MAE 0.06 m for both original FPD-filtered Displacement dune crest lines ALS resulted migration rate ≈ 5 m/year between 2010 2019, agreement rates derived satellite historical aerial photographs area. volume change period showed decrease only 0.2%, which can be related installation fences promote stabilization removal front field keep road open vehicles. cover large little time its high cost still remains barrier wider usage, especially researchers developing countries. has intermediate demands more processing time. our case needed three days around weeks produce 80 400m2. On other hand, were able 740 900m2 six flight missions under hours, 13 h medium-range workstation. This makes low-cost solution fast reliable 3D modelling continuous monitoring dunes.