#3 Multi-stereo Satellite Images data processing for Digital Surface Modeling
High resolution (High-res) satellite images nowadays can observe the ground as small as a 30-cm footprint (e.g. worldview 3/4 sensors), and play an important role for wide-area 3D data generation. This tutorial provides both theory and practice session focusing on the use of photogrammetric techniques for generating 3D digital surface models. The theory session will provide an introduction to the basic concepts in satellite based mapping, as well as our best practices in processing both on-track and incidental satellite images. It also includes a brief introduction to the processing components such as geo-referencing, efficient surface reconstruction using empirical and deep-learning based methods. The practice session will provide the attendants multi-view satellite datasets and necessary software to process them. Specifically, the RPC stereo processor (RSP) that provides both empirical and deep-learning based matcher will be used to practice all stages of satellite image processing, including level-1 data correction, pan-sharpen, geo-referencing, dense image matching, surface reconstruction and orthophoto rectification. RSP as a core of solutions have won a few international contests including IARPA multi-view 3D challenge and IEEE Data Fusion contest 2019. This tutorial targets on students, faculty, researchers and working professions in data collection, mapping and urban/environmental applications.