Vegetation Height Stereo Reconstruction With BlackSky Commercial Frame Camera Imagery

Commercial stereo imagery has provided unprecedented planar detail (< 2 m) of the Earth’s surface. However, a number compounding factors, including view angle, convergence angle, imaging detector design, ground sampling distance (GSD), time of day/year etc., induce bias in surface reconstructions. BlackSky 1 m GSD data and its unique frame detector/telescope configuration provides a novel resource with sensor capabilities of staring, and motion imagery/video to benchmark existing commercial digital elevation model (DEM) products. The BlackSky constellation with its high repeat collection capability also allows for dense imagery collection and enables multi-view stereo reconstruction. We evaluated the ability of BlackSky data for vegetation surface reconstruction using a multi-view stereo methodology and compared products to airborne LiDAR data. We found an Absolute Median Error of 1.77 m and normalized mean absolute deviation (NMAD) of 1.54 m in areas with sufficient volume of available digital surface models (DSMs). Our results indicate the potential benefit of this workflow and resulting products for vegetation analyses.