An Efficient and Robust Singular Value Method for Star Pattern Recognition and Attitude Determination

A new star pattern recognition method is developed using singular value decomposition of a measured unit column vector matrix in a measurement frame and the corresponding cataloged vector matrix in a reference frame. It is shown that singular values and right singular vectors are invariant with respect to coordinate transformation and robust under uncertainty. One advantage of singular value comparison is that a pairing process for individual measured and cataloged stars is not necessary, and the attitude estimation and pattern recognition process are not separated. An associated method for mission catalog design is introduced and simulation results are presented. Introduction Star cameras are among the most attractive attitude sensors because they provide three-axis attitude information with high accuracy. Star pattern recognition is an integrated part of star cameras. It is based on the invariant properties from coordinate transformation. For the past two decades, many of the invariant properties, such as angular separation, brightness of star, shape of triangles with vertices of star, and constellations, have been utilized for star identification. In practice, one of the challenges of using these star trackers arises when initializing sensors and/or recovering them from sudden failure is to efficiently solve the “lost-in-space” problem. Even though a-priori attitude estimation is not necessary for a star tracker, these initializing problems without a-priori attitude information tend to require intensive computer storage and 1 Principal Scientist, Structural Dynamics 2 Graduate Student, Department of Aerospace Engineering 3 Distinguish University Professor, Department of Aerospace Engineering