Star Identification Algorithm Based on K–L Transformation and Star Walk Formation

For star sensors, the difficulty in realizing attitude determination lies in accurate and robust star identification algorithms. In this paper, an efficient star identification algorithm is described, which is fast, robust, and consumes small memory resources. It mainly involves two steps: the dedicated Karhunen-Loève (K-L) transformation step and the star walk formation step. Once several pattern triangles for pole stars are built, the K-L transformation is performed to calculate the K-L values to match several rough correspondences from a predefined database. Star walks are then formed among the candidate stars using the neighboring star table, which indicates each star's geometric distribution in the neighborhood, until a unique match is found. Our algorithm can also complete an efficient verification procedure based on a fast triangle algorithm, since stars' surrounding information can be acquired from the neighboring star table. The simulation results compare favorably with two published algorithms, and the approach has been implemented on actual star images.