Lunar Terrain Depth Mapping and 3D Reconstruction: Benchmarking Seven Depth Estimation Models

This paper evaluates seven advanced depth estimation frameworks-GLPN, MiDaS, Stereo Matching, DPT, Zoedepth, Depth Anything, and Marigold-for lunar surface mapping and 3D reconstruction. It focuses on key factors such as edge retention, texture fidelity, depth reliability, and geometric resolution. The analysis highlights each model's strengths and limitations in capturing the complexities of the lunar landscape, offering insights into model selection for lunar missions and planetary research. The findings aim to enhance the precision of lunar terrain mapping and improve the accuracy of 3D reconstructions.