Single image based hazard detection for a planetary lander

Autonomously landing a spacecraft safety in a hazardous area on a planet is a critical mission goal. To help achieve this goal, we developed a fast algorithm to detect hazards on a planetary surface using a single image from a descent spacecraft camera. This paper proposes a general approach of segmentation and classification for hazard detection and describes all implementation of this approach. The implemented algorithm uses local intensity clustering to segment the image into regions and then classifies these regions as hazards or part of the surface. We tested our methodology using synthetic terrain images as well as real images.