Detection and Initial Assessment of Lunar Landing Sites Using Neural Networks

TL;DR

This paper presents a neural network-based system using MobileNetV2 and monocular structure from motion to autonomously detect and assess safe lunar landing sites, enhancing precision and safety for future lunar missions.

Contribution

It introduces a passive hazard detection system combining neural networks and surface reconstruction for initial landing site assessment.

Findings

Successful detection of hazards like rocks, shadows, craters

Effective surface slope and roughness analysis from monocular images

Potential for autonomous, real-time landing site evaluation

Abstract

Robotic and human lunar landings are a focus of future NASA missions. Precision landing capabilities are vital to guarantee the success of the mission, and the safety of the lander and crew. During the approach to the surface there are multiple challenges associated with Hazard Relative Navigation to ensure safe landings. This paper will focus on a passive autonomous hazard detection and avoidance sub-system to generate an initial assessment of possible landing regions for the guidance system. The system uses a single camera and the MobileNetV2 neural network architecture to detect and discern between safe landing sites and hazards such as rocks, shadows, and craters. Then a monocular structure from motion will recreate the surface to provide slope and roughness analysis.