Applications of High-Resolution, Cross-Track, Pushbroom Satellite Images With the SETSM Algorithm

摘要

Digital elevation models (DEMs) and surface displacement maps (SDMs) obtained from repeat satellite imagery provide critical measurements of changes in the earth's surface over large, remote areas. While DEMs are typically extracted from stereo pairs obtained along the same orbital pass (i.e., in-track stereo), single acquisitions are more abundant and stereo pairs created from repeat-single scan images offer the potential to increase the temporal and spatial coverage of DEMs. If precisely coregistered, repeat cross-track images may also provide measurements of surface displacement through feature tracking. We present a methodology for obtaining both DEMs and SDMs from sequences of repeat, submeter resolution, pushbroom satellite images that build upon the framework of the Surface Extraction from TIN-based Search-space Minimization algorithm. We first demonstrate a Local Surface Fitting (LSF) for reducing the noise in DEMs caused by narrow convergence angles between cross-track image stereo pairs. We then detail a procedure in which a reference image stereo pair is combined with a third repeat image to simultaneously extract surface elevation and displacement, providing DEMs and SDMs that are precisely coregistered and optimally terrain corrected. We conclude with example applications to a fjord with moving icebergs and a fast-flowing glacier and an assessment of DEM quality assessment and improvement by the LSF.