Redundant-coded radial and neighbor star pattern identification algorithm

Abstract

Aimed at solving the “lost-in-space” (LIS) problem, a robust star pattern identification algorithm is presented. The neighbor star pattern is introduced to increase separating ability and stabilization of star patterns and reduce the effects of magnitude noise. Redundant coding for radial and neighbor star patterns is proposed to more effectively remove the effects of position noise. Binary bit strings are used for storing the redundant code to reduce the size of the on-board database. A similar score measurement is proposed on the basis of shift and logical AND operations to quickly implement the star's initial match. The final match results are obtained using a strategy of searching for the longest matching chain. The algorithm is evaluated with synthesized and real star images. The identification rates of this algorithm are 97.98% with standard deviations of position noises of 1 pixel and 94.96% with standard deviations of magnitude noises of 0.646 Mv. Moreover, the identification rate is 77.12% when there are only four stars in the field of view (FOV). The algorithm obtains an average identification rate of 96.96% from real star images.