A Polygon-Based Joint Feature Construction and Matching Algorithm for Star Identification in Star Sensors

To address the insufficient discriminative capability of features and the susceptibility to mismatches under noise in conventional geometric-feature-based star identification algorithms, this paper proposes a polygon-based star identification algorithm that integrates multiple features. The algorithm combines inter-star angular distances, angles between adjacent edges, relative stellar magnitude differences, and multi-order similarity invariants to construct joint polygon features. The Mahalanobis distance is used for matching decisions, thresholds are determined from the chi-square distribution, and an optimized retrieval strategy prioritizing five-star polygons and supplementing them with four-star polygons is designed. Using 372 measured star images as the basis for 11 test conditions, the proposed algorithm is compared with the triangle, pyramid, and conventional geometric polygon algorithms on a frame-by-frame basis, with each algorithm completing 4092 frame-wise solutions. The results show that the proposed algorithm achieves the highest identification rate among the four algorithms under all non-photometric perturbation conditions, ranging from 84.14% to 92.74%; its identification rates under 1.0 px star position noise and 20% star dropout are 92.20% and 86.02%, respectively, both exceeding those of the three comparison algorithms. Under all-sky lost-in-space conditions, the 95th percentile of the runtime for correctly identified frames is below 1 s in every environment, and the mean solution runtime is reduced by 41.5%, 94.6%, and 82.1%, respectively, relative to the three algorithms above. It exhibits good noise tolerance and identification efficiency and can provide methodological support for all-sky lost-in-space star identification.