REFINEMENT OF STEREO IMAGE ANALYSIS USING PHOTOMETRIC SHAPE RECOVERY AS AN ALTERNATIVE TO BUNDLE ADJUSTMENT

Abstract

Topographic mapping, e.g. the generation of Digital Elevation Models (DEM), is of general interest to the remote sensing community and scientific research. Commonly, photogrammetric methods, e.g. stereo image analysis methods (SIAM) or bundle adjustment methods (BAM), are applied to derive 3D information based on multiple images of an area. These methods require the detection of control points, i.e. common points within multiple images, which relies on a similarity measure and usually yields a sparse map of 3D points. The full spatial DEM is then obtained by interpolation techniques or imposed restrictions, e.g. smoothness constraints. Since BAM utilizes all images of the area, it is assumed to provide a more accurate DEM than SIAM which utilizes only pairs of images. Intensity-based shape recovery, e.g. shape from shading (SfS), utilizes the reflectance behavior of the object surface and thus provides a dense map of relative height changes, which provide the possibility to refine the photogrammetric DEMs. Based on Rosetta NavCam images of 67P/Churyumov-Gerasimenko we compare intensity-based DEM refinement methods which use DEMs obtained based on SIAM and BAM as a reference. We show that both the SIAM based DEM refinement and the BAM based DEM refinement are of similar quality. It is thus possible to derive DEMs of high lateral resolution by applying the intensity-based refinement to the less complex SIAM.