Artificial Intelligence Assisted Crater Detection for Lunar Surface Landing and Terrain Relative Navigation

Lunar surface landing and terrain navigation is an essential component of space exploration missions. In such missions, accurate detection of craters is critical for a safe landing and efficient navigation. To this end, this study focuses on the problem of crater detection in lunar surface images acquired at different distances and resolutions. Three different datasets consisting of images obtained at 5 $\mathrm{km}, 10 \mathrm{~km}$ and 20 km scales were used in the experiments. YOLO and DETR architectures for crater detection are investigated and their performance is compared. The results show that the models successfully detect craters with mAP50 scores of $80 \%$ and above. In this context, it is concluded that the proposed models can help the spacecraft to successfully detect and avoid craters at every stage of its approach to the lunar surface. This study sheds light on the development of crater detection techniques for future space exploration missions and lunar surface landing projects.