Crater detection, classification and contextual information extraction in lunar images using a novel algorithm

This study presents the development and implementation of an algorithm for automatic detection, classification and contextual information such as ejecta and the status of degradation of the lunar craters using SELENE panchromatic images. This algorithm works by a three-step process; first, the algorithm detects the simple lunar craters and classifies them into round/flat-floor using the structural profile pattern. Second, it extracts contextual information (ejecta) and notifies their presence if any, and associates it to the corresponding crater using the role of adjacency rule and the Markov random field theory. Finally, the algorithm examines each of the detected craters and assesses its state of degradation using the intensity variation over the crater edge. We applied the algorithm to 16 technically demanding test sites, which were chosen in a manner to represent all possible lunar surface conditions. Crater detection algorithm evaluation was carried out by means of manual analysis for their accuracy in detection, classification, ejecta and degraded-state identification along with a detailed qualitative assessment. The manual analysis depicts that the results are in agreement with the detection, while the overall statistical results reveal the detection performance as: Q ∼ 75% and precision ∼0.83. The results of detection and classification reveal that the simple lunar craters are dominated by the round-floor type rather than flat-floor type. In addition, the results also depicts that the lunar surface is predominant with sub-kilometer craters of lesser depth.