Autonomous Craters Detection from Planetary Image

As development of deep space exploration, the guidance, navigation and control (GNC) technology of spacecraft or probe is becoming more important than ever. Vision-based navigation (optical navigation) is a good method to achieve autonomous landing of spacecraft. Therefore, the landmark has to be detected for vision-based navigation. Craters are commonly found on the surface of planets, satellites, asteroids and other solar system bodies. Currently, the most of optical landmark navigation algorithm are built on the craters detection and tracking. The focus of this paper is to present an algorithm for autonomous crater detection. The whole course of crater detection can be divided into two steps, multi-resolution feature points extraction and crater detection. The first step can be further divided into multi-resolution window-based feature points' extraction and crater candidate area choice. The second step can be further divided into region growing, pixels of crater edge extraction, ellipse detection and obtaining craters. Experimental studies demonstrate that the detection rate of this algorithm is higher than 90% for image where the distribution of craters is discrete.