Abstract: | There are two difficulties in star spot extraction in high dynamic conditions. Firstly, dim and small stars are difficult to identify in noise, and the accuracy of star centroid is poor. Secondly, the star spots may be damaged, which cannot be extracted by common domain algorithm. In view of the above difficulties, the star spot model under high dynamic conditions was established. The star spot compensation and centroid location algorithm was proposed based on the model. The algorithm includes four steps. Firstly, Kalman filter was used to iterate the static model of star spot in real time. Secondly, the dynamic model of the star spot was established based on the static model and the motion blur model. Thirdly, the dynamic model, as a template for correlation matching in the tracking window, was used for the position determination of star spot. Fourthly, the star spot was compensated based on the dynamic model, and the star centroid was calculated. Simulation results showed that the algorithm can effectively solve the problem of dim mall star spot extraction and damaged star spot restoration in high dynamic conditions. Compared with the traditional algorithm, the mean error of attitude accuracy is reduced by 40.9%, the maximum error is reduced by 81.2%, the star extraction rate is 100%, the star extraction rate is increased by 174.5%, and the threshold segmentation is improved by 173% compared with the combined domain method. |