Cornell University) on August 2, 2025. It has been sent to the Institute of Electrical and Electronics Engineers (IEEE) for publication.
The study addresses the increasing complexity of planetary missions that currently depend on digital terrain maps (DTMs) for landing and navigation in GPS-denied environments like Mars. However, DTMs are challenging to utilize, requiring manual image refinement and having issues with variable lighting. The ongoing exploration of locations such as Mars hinges on the availability of high-quality DTMs. The researchers explored AI methods like neural volume rendering to tackle this problem.
Their proposition is known as Neural Terrain Maps (NTM). It integrates data into comprehensive 3D maps derived from satellite images using cutting-edge AI rendering techniques. These maps will boast high resolution, even when using imperfect camera data. This approach may enhance future Mars missions, for instance, in investigating the critical possibility of water on Mars.
