月球探测器鲁棒环形山检测及光学导航方法

Traditional crater detection algorithms (CDAs) are photosensitive and have poor robustness. To solve this problem, a new CDA based on maximum entropy threshold ternary segmentation is proposed and used in optical navigation. This method uses different filter kernel functions to eliminate noise and thereby smooth an image. Furthermore, the processed image is segmented by the maximum entropy threshold, and the image information is trivalued to remove the photosensitivity of the image while retaining the image information to the greatest extent. A normalized multi-indicator constrained crater matching and a ellipse fitting method are proposed for complete crater extraction. This CDA method is applied to optical navigation Monte Carlo experiments and verify the real-time algorithm. Simulation results show that compared with traditional methods based on morphology or adaptive edge detection, the proposed CDA can extract continuous and smooth crater edges on a large scale, increasing the number of effective craters by more than 35% while reducing the calculation cost by 40%. The optical navigation algorithm based on robust crater detection algorithm has better real-time performance.