Automatic detection of lunar craters based on DEM data with the terrain analysis method

Impact craters are the most obvious tectonic landform unit on the lunar surface and are highly significant in the study of lunar geomorphic features. Currently, the mainstream automatic detection algorithms used to identify lunar impact craters represent the craters as circles, which does not accurately reflect their true shapes, making it difficult to measure characteristics such as circularity and diameter. In this study, morphology of lunar craters were analyzed based on 100 m resolution digital elevation models (DEM), and a new crater detection algorithm (CDA) is proposed by extracting higher change rate of slope of aspect (SOA) values at crater rims. Then, the neighborhood mean algorithm and reclassification method are used to obtain the positive terrain, which can filter the noises of non-crater rims. Finally, complete and true impact crater boundaries are obtained using a morphological treatment, and further denoising is simultaneously conducted. The results of experiments in the D'Alembert (lunar highland) and Serenitatis (lunar mare) areas show the following: Compared with the LU60645 crater catalog, the factors for our algorithm were: detection percentage (D) were 89.9% and 89.5%, branching factor (B) were 0.1 and 0.05, and quality percentage (Q) were 82.2% and 85.9%, respectively. Compared with the other algorithms, the Q and the B in the proposed CDA are better than those. It shows that the proposed algorithm maintains a relatively low false detection rate while maintaining a higher correct detection rate. Moreover, the extraction results closely correspond to the natural forms.