A star shortlisting technique for a lost-in-space mode star tracker

A star tracker has become the most extensively used sensor for attitude determination of a satellite in the past decade. It normally operates in two modes: lost-in-space (LIS) mode, when no prior attitude information is available and once the initial attitude is acquired, it shifts to the tracking mode. Star pattern recognition (SPR) technique forms the most crucial part of a star tracker in the LIS mode. An SPR technique is evaluated on two parameters — recognition reliability and time of recognition. In this paper, we propose a novel star shortlisting technique to shortlist a part of the star pattern database (SPD) i.e. the star IDs which moderately seem to be present in the image. Later, these shortlisted star IDs can be supplied to the existing SPR techniques, thus allowing the SPR techniques to operate only on a part of the SPD which will lead to a reduction in time for these SPR techniques. The proposed approach is based on constructing a rotation and translation invariant vector frame, and later comparing the vectors constructed in this new frame with the ones stored in the SPD. The proposed star shortlisting technique can shortlist one-third of SPD with a high probability of 0.99 of containing the correct star ID. When combined with the existing SPR techniques, the corresponding time reduction for the SPR techniques is one-half of their original time.