Impact Crater Databaseof 10 Apollo and Chang'e Landing Regions: Deep Learning Driven Construction and Distribution Patterns of Over 350000 Craters with

Impact craters are key elements in interpreting the orbital evolution of solar system bodies and the surface age of terrestrial planets, and the Moon is the best-preserved inner solar system body in terms of impact craters. Impact processes are one of the most important geological processes in reshaping airless bodies, and lunar landing sites with in-situ detection data and returned samples are important research objects for understanding impact events and their reshaping effects. Existing lunar crater databases lack sufficient coverage of small-scale craters with diameters less than 100 meters, yet these small craters play a significant role in the formation and evolution of lunar regolith. This study constructs a database of craters with diameters greater than 15 meters across 10 landing sites from six Apollo missions and four Chang'e missions. This paper first proposes the YOLO11+SAHI fusion model, which integrates the cross-stage kernel optimization module (C3K2) to enhance edge feature extraction, uses the spatial pyramid fast pooling module (SPFF) to address scale sensitivity, employs the parallel spatial attention module (C2PSA) to suppress false positives in complex terrain, and combines the slice-assisted super-inference (SAHI) strategy to enhance small target recognition capabilities. The final model achieved an average precision of 0.985, a recall rate of 0.94, and an F1-score of 0.962 on the test set, significantly improving the robustness of small-scale impact crater detection. Based on LROC NAC high-resolution imagery, the model was used to automatically extract craters within a 2,020 km diameter range of 10 landing sites. After manual verification, a crater database containing 357,764 records was established. Based on crater density distribution and size-frequency distribution characteristics, it was revealed that secondary craters have a significant impact on the distribution of small-sized craters. Compared to previous studies, the database has significantly improved the completeness of the database and the proportion of craters within the complete range. The database established in this study provides important research data for artificial intelligence models related to craters, and can also provide important support for lunar landing site geological age calibration, impact flux evolution, lunar surface geological evolution, and returned sample research.