Unified redundant patterns for star identification

Star pattern identification algorithms have employed widely to identify observed stars by star trackers which require reliabilities and adaptivities of algorithms. However, the noises especially the number of stars in field of view (FOV) are still serious problems for primitive star pattern algorithms. A novel algorithm of unified redundant patterns, which are combined with novel redundant neighbor patterns and redundant radial patterns, for star trackers in the “lost in space” mode is presented. Since neighbor patterns and radial patterns describe two different distributions, the distributions between neighbor stars each other and the distributions in the radial direction respectively, and constructed with similar redundant coding solution, they are combined to unified redundant patterns and stored by binary bit strings which reduce memory requirement of on-board database significantly. For the pattern redundancies and combinations, as well as similar score measurement applying patterns synthetically other than multiple-step match, consequently, the proposed algorithm is robust to the star positional noise, magnitude noise and performs still well when sparse stars are in FOV. When evaluated with synthesized star images, our algorithm can obtain identification rates of 99.14% (positional noise is 0.4 pixels). When there are only 4 stars in FOV, the algorithm can still obtain identification rate of 77.03%, much higher than the other star pattern identification algorithms.