Automated digital elevation model (DEM) generation from very-high-resolution Planet SkySat triplet stereo and video imagery

摘要

The Planet SkySat-C SmallSat constellation can acquire very high resolution (0.7 m to 0.9 m) triplet stereo and video imagery with short revisit times, providing an exciting opportunity for global, on-demand 3D mapping of dynamic surface features. However, a lack of suitable processing software, limited geolocation accuracy, and scene-to-scene offsets currently limit the potential for accurate SkySat digital elevation model (DEM) production. We developed an open-source workflow to refine the SkySat-C camera models and improve absolute image geolocation using external reference DEMs, without manual ground control point (GCP) selection. The refined camera models are used to generate accurate and self-consistent DEMs with 2-m posting and orthoimages at native resolution. We present sample DEM products for a triplet stereo collection over Mt. Rainier, USA and two video collections over Mt. St. Helen's, USA. The output DEMs display 1 to 2 m relative and 2 to 3 m absolute vertical accuracy when compared to DEMs generated with stereo image pairs acquired by the DigitalGlobe/YMaxar WorldView satellites and airborne LiDAR. Differencing the two SkySat-C video DEMs over Mt. St. Helen's shows elevation change of similar to 5 to 15 m due to melting of seasonal snow and glacier flow. Our workflow can be scaled for batch processing of SkySat stereo imagery, and extended to other frame camera systems with limited initial geolocation accuracy.