Construction of Digital Terrain Maps From Multiview Satellite Imagery Using Neural Volume Rendering

Digital terrain maps (DTMs) are an important part of planetary exploration, enabling operations such as terrain relative navigation during entry, descent, and landing for spacecraft and aiding in navigation on the planetary surface. As robotic exploration missions become more ambitious, the need for high-quality DTMs will only increase. However, producing DTMs via multi-view stereo pipelines for satellite imagery, the current state-of-the-art, can be cumbersome and require significant manual image preprocessing to produce satisfactory results. In this work, these shortcomings are addressed by adapting neural volume rendering techniques to learn textured digital terrain maps directly from satellite imagery. The proposed method, neural terrain maps (NTM), only requires the locus for each image pixel and does not rely on depth or any other structural priors. NTM is demonstrated on both synthetic and real satellite data from Earth and Mars encompassing scenes on the order of 100km2. The accuracy and precision of the output terrain maps is evaluated by comparing with existing high-quality DTMs produced using traditional multi-view stereo pipelines. NTM shows promising results, with the precision of terrain prediction almost equal to the resolution of the satellite images even in the presence of imperfect camera intrinsics and extrinsics. As compared to high quality reference terrain maps, NTM results ranged from an error standard deviation of 498m for ASTER imagery with a grounds sample distance (GSD) of 15m to 19m for MRO CTX imagery with a GSD of 6m.