Rapid Single Image-Based DTM Estimation from ExoMars TGO CaSSIS Images Using Generative Adversarial U-Nets

The lack of adequate stereo coverage and where available, lengthy processing time, various artefacts, unsatisfactory quality complexity automating the selection best set parameters, have long been big barriers for large-area planetary 3D mapping. In this paper, we propose a deep learning-based solution, called MADNet (Multi-scale generative Adversarial u-net with Dense convolutional up-projection blocks), that avoids or resolves all above issues. We demonstrate wide applicability technique ExoMars Trace Gas Orbiter Colour Stereo Surface Imaging System (CaSSIS) 4.6 m/pixel images on Mars. Only single input image coarse global reference are required, without knowing any camera models imaging to produce high-quality high-resolution full-strip Digital Terrain Models (DTMs) in few seconds. discuss technical details system provide detailed comparisons assessments results. resultant 8 CaSSIS DTMs qualitatively very similar 1 HiRISE DTMs. display excellent agreement nested Mars Reconnaissance Context Camera (CTX), Express’s High-Resolution (HRSC), Laser Altimeter (MOLA) at large-scale, meanwhile, show fairly good correlation Science Experiment (HiRISE) fine-scale details. addition, how outperforms traditional photogrammetric methods, both speed quality, other datasets like HRSC, CTX, HiRISE, parameter tuning re-training model. results Oxia Planum (the landing site European Space Agency’s Rosalind Franklin rover 2023) couple sites high scientific interest.