FOSS4G DATE for DSMs generation from tri-stereo optical satellite images: development and first results

Abstract

The topic of this research is the identification of an innovative strategy for DSMs generation from optical satellite tri-stereo imagery, exploiting efficient dense matching algorithms from computer vision, without losing a rigorous photogrammetric approach. The main challenge is related to the epipolarity resampling for satellite images, for which it is not straightforward the epipolar geometry achievement, due to their multiple projection centres. The Ground quasi-Epipolar Images (GrEI) are the core of this original strategy for DSMs generation, able to return reliable and accurate solutions. The FOSS4G DATE has been modified since its original implementation, in order to obtain better results in terms of DSMs generated, and a new version has been released. The workflow provides a ground projection for the raw satellite images, exploiting RPFs model and a coarse DSM. In order to bring the ground images in the quasi-epipolar geometry, a preliminary rotation is estimated for both images to align them in the disparity prevailing direction. Then, a refinement of their relative position is performed. These orthorectified images can act as GrEI, and can undergo a dense image matching procedure. Tests have been carried out with Pléiades triplets over Trento (Northern Italy) and very good results are achieved.