Deep learning for lunar crater detection

Abstract

Since Apollo 11 landed on the moon in 1969, the human civilization made significant progress in understanding our closest astronomical object - The Moon. Scientists have been fascinated with the lunar surface as it represents our Solar System’s violent past. The fact that the Moon has no atmosphere renders it prone to asteroid impacts, and with negligible geological activity on the Moon, those asteroid impacts have been preserved for centuries. The craters on the Moon give us a glimpse of what has happened in the geological past in our immediate vicinity, and hence, scientists have been trying various methods to find and map lunar craters to understand their physical characteristics. This thesis is aimed to analyze the current techniques used to detect craters on the lunar surface. This thesis is specifically aimed to study deep learning based methods and contrast them to the vanilla statistical and human-enabled methods to detect craters. Finally, this thesis also presents an innovative approach to improve on currently used techniques for lunar crater detection.