RPC-based adjustment model for COSMO-SkyMed stereo slant/ground-range images

Abstract

The use of stereoscopic SAR images offers an alternative to conventional stereo-photogrammetric survey for the generation of Digital Elevation Models (DEMs). Often the SAR ground-range form is more popular with the commercial users, since the pixel spacing on the ground is roughly the same for the different look-angle images. The different mathematical descriptions between the slant-range and ground-range products thus make the stereo modeling and adjustment a challenging objective to deal with. Previous work applied sensor model adjustment of range and timing parameters to SAR spotlight range images, promising 3-D mapping accuracies in the range of 2 m. However, this adopted the direct least squares method, which is too sensitive in the case of the ground-range images, making it less than optimal for SAR range images' stereo restitution. In this paper, an image based transformation (geometric correction) using a small number of control points(CPs) in cooperation with the Rational Polynomial Coefficient (RPC) model to improve the space intersection accuracy is proposed. The development of such an RPC-based adjustment method is first described, which is very practical to implement and can be applied to SAR slant- or ground-range products. By situating several well-distributed trihedral corner reflectors (CRs) within sites and imaging these sites using COSMO-SkyMed's stripmap (SM) mode, the modeling quality of the delivered slant- or ground-range products was validated, and the 3-D mapping potential was also assessed.