Automated Lunar Crater Mapping with Custom CNN

The surface of the Moon holds valuable clues about its formation and the history of our solar system. Craters, formed by billions of years of meteor impacts, help scientists understand the age, structure, and composition of lunar regions. However, identifying and mapping craters manually from satellite images is slow, error-prone, and limited in scale. This research presents an automated approach using deep learning to detect and classify lunar craters from high-resolution images captured by Chandrayaan-2’s Terrain Mapping Camera-2 (TMC-2). We trained a Convolutional Neural Network (CNN) using TMC-2 data to identify the main features of craters, including the circular rim and the shadows as well as the patterns in the event of ejecta. Our aim is to develop a quick and reliable method of crater identification that minimizes human labor efforts while being more accurate. This method of automation could contribute to future lunar missions by producing detailed maps on the lunar surface assisting in target site selection.