A Robust Subpixel-Level Celestial Edge Localization Algorithm for Optical Navigation

The precise localization of celestial edges is crucial for autonomous navigation in deep space. This study proposes a high-precision subpixel-level edge localization method based on coarse and fine detection techniques for interplanetary optical navigation systems. First, pixel-level detection is improved by integrating the 8-neighborhood edge gradient and subdividing the edge width into a single pixel. Second, an accurate subpixel detection method using Zernike moments with the ramp edge model is advanced after analyzing the celestial object’s light intensity distribution and edge intensity. Third, the Otsu algorithm is integrated into the Zernike moments’ method, effectively improving the efficiency of the threshold selection. Finally, experiments conducted on simulated and real images, in comparison with three representative algorithms, show that the proposed algorithm can accurately estimate subpixel edges of celestial bodies while maintaining good robustness.