Autonomous Star Identification using Pattern Code

In the work presented here, a new star identification strategy is proposed. In the proposed method the pattern of nearest k neighbor stars of each star is coded into a translation, rotation, and scale-invariant integer. If the code words of any two stars have equal integer values, it implies that the patterns of their nearest k neighbors are similar. Therefore, a database search for a (k+1)-polygonal pattern can be equated to a simple search for an integer value. This simplified method significantly reduces the complexities associated with star identification. By using the code words as indexes for the lookup table, the time required to solve the star identification problem is determined based on the number of stars within the field of view (FOV) and not the number of catalog stars. Simulations show that the proposed pattern-coding algorithm is robust and efficient. The developed algorithm has successfully identified stars from night sky photographs of 0.21°×0.21° FOV.