Image Processing-Based Real-Time Safe Site Identification for Autonomous Landing

Abstract

Autonomous hazard detection and avoidance (HDA) is essential in both uncrewed aerial vehicle (UAV) and planetary landings, enabling successful landings despite potential hazards unrealized during mission planning. Various methods are proposed for safe landing site detection, but computational complexity limits their onboard real-time application. This paper introduces a real-time method to identify potential sites for safe autonomous landing based on image processing. The approach adopted in this paper regards the digital elevation model (DEM) as a two-dimensional array rather than geographic information. It employs image processing techniques using kernels to ensure real-time computation. Three safety maps were generated to evaluate the safety of the DEM for different purposes. The weighted sum optimization of generated safety maps is used to select the final landing site. The proposed method is evaluated by conducting two distinct case studies, encompassing UAV landing and Mars landing scenarios.