A fast star pattern recognition algorithm based on feature vector

Abstract

Stars in a star map can be regarded as a point pattern, and we can utilize the matching of point pattern to recognize the star pattern. First, the nth Radius-Weighted-Mean Points (RWMPs) are proposed which are invariant to translation, rotation and scaling, and then, a RWMP-based feature vector is constructed which is still invariant to translation and rotation. The candidate referenced star images and their corresponding attitudes are obtained by computing the Euclidean distance between the viewed star image and each of the star images in the pattern database. The verification process is introduced to confirm the identification results. The simulation results indicates that the average identification rate of this algorithm can be enhanced 3.5% as compared to the grid algorithm at the same position noise level from 0 to 3 pixels, and the identification time of the proposed algorithm reduces to 1/5.