Rock detection on mars via superpixels method

A novel rock detection method on Mars is proposed via superpixels in this paper. Automatic hazards detection remains an interesting yet challenging topic in the past decade. It plays a crucial role whether in the initial landing site assessment and selection for the mission or in the descent phrase of the lander, even in the route planning and geologic analysis for Mars rover. Using the images of planar surface, hazards posed by rocks can be detected by some auto rock detection algorithms. Current state-of-art detection algorithms are usually time-consuming to be practical or imprecise in adhering to rock boundaries. Accurate and real-time rock detection would guarantee the success of the mission. SLIC superpixels method is a novel faster and more memory efficient algorithm than previous methods. The speed advantage makes SLIC capable to handle large image which contains more information with small size rocks. Multi-scale features can reduce the detection errors as a result of directional lighting, variable rock morphology and weak local cues. New method utilized multi-scale detectors to describe the appearance of rocks and preliminary results show that it may be practically applied to rock detection process in future Mars missions.