Effect of sparse recovery on centroid estimation for star image

The compressive sensing (CS) theory has represented an attractive perspective that can effectively reduce the sampling rate of star tracker. The star centroid is the key factor relating the star identification and attitude determination. Naturally, there exists a problem that whether the low sampling rate of star tracker can keep the star centroid as its original position to perform the successful star identification. In order to address the problem, this paper analyzes the effect of recovery on star centroid estimation in the sparse representation field. Particularly, a concept of successful rate of star recovery is presented to evaluate the quality of sparse representation on star point. Furthermore, the successful rate is investigated with regard to the following aspects: the sparse representation method, the sparsity and the visual magnitude of stars. The numerical simulation shows that the sparsely recovered star point can keep its centroid to a large extent and meet the requirement of successful star identification, which proves the feasibility of applying CS theory to star tracker.