Optical Detection of Lunar Surface Obstacles

Abstract

Space missions aiming to land on celestial bodies like the Moon must be very careful to avoid surface obstacles like craters or large rock formations; contact with such objects would likely spell disaster for the mission. Landing site selection has often been a slow and tedious process for scientists, requiring detailed human analysis of surface topography. This thesis aims to explore a supplementation or alternative to human-selected landing site determination; use of an on-board camera to capture images of surfaces and quickly identify rocks and craters to avoid. Sample image data, captured by the Lunar Reconnaissance Orbiter, was used to test a digital image processing algorithm’s ability to isolate and identify such dangerous objects of interest in a rapid, autonomous manner. These identified objects’ locations were then stored and tracked over time, providing a theorized spacecraft with the ability to precisely determine which regions of a body’s surface are safe to land on at any given time. The idea of utilizing object locations as a way of determining the dynamics of the spacecraft – using optical flow to estimate velocity and acceleration, specifically – is briefly explored as well. With future work, the ultimate goal would be to enable a camera to provide equivalent data to a number of sensors currently in use on space missions, while simultaneously speeding up the landing site selection process and enabling automatic determination of surface safe zones. With an object detection accuracy of 87% and a proven ability to algorithmically select suitable landing sites given input image data, this thesis demonstrates a strong proof of concept of the explored methods.