A Neural Network Star Identification Algorithm Based on Equal-Frequency Binning Radial Feature

Abstract

Star identification algorithm is the core algorithm of star sensor. The number of false stars in the star image taken by the star sensor with wide Field-Of-View will increase, and the position of star will be seriously offset. This paper proposes a neural network star identification algorithm based on equal-frequency binning radial feature (EFB-RF). The EFB-RF is a robust star pattern. A superficial neural network is used to classify EFB-RF. The results show that our algorithm is robust to position noise, false stars, and missing stars. The identification rate of our algorithm can reach 99.92% under 1 pixel position noise, and it can reach 99.37% under 5% false stars. When the percentage of missing stars is 15%, it can reach 99.97%. The results show this algorithm is robust to position noise, false star and missing star in a wide-FOV, and it can help star sensor work properly in complex space environment.