An Efficient Guide Stars Classification Algorithm via Support Vector Machines

The purpose of this study is to obtain an approximate even guide stars catalog (GSC) applied in star trackers, thus a guide stars selection algorithm via support vector machines (SVM) is presented. Using combination of the number of stars and Boltzmann entropy within circular region centered at every star of original catalog(OC) as feature vector, the local density and uniformity of each star from OC is characterized preferably, which distinguishes guide stars and non-guide stars meeting structural risk minimization (SRM). The SVM algorithm is implemented by generating the GSC for a star tracker with an 8degtimes8deg squared field of view (FOV). To validate the GSC generated by SVM, statistics of guide stars number inside the FOV is compared between SVM and magnitude filtering method(MFM) using 10,000 random boresight directions. Results clearly show the volume of GSC created by the SVM algorithm is approximately 34% and the standard deviation is 22% accounting for that of MFM satisfying four guide stars inside the FOV. Consequently, the proposed algorithm makes a great progress relative to MFM in capacity and uniformity of GSC.