A New Sub-Pixel Subdivision Location Algorithm for Star Image

A new sub-pixel subdivision location algorithm for estimating the position of the star on the focal plane array is proposed. An analytical study of the optics theoretical point spread functions is presented and it is shown that the star intensity distribution is close to the Gaussian point spread function model. Based on the numerical model of the star, a Gaussian curved fitting algorithm for centroiding location of the star image is devised. Since behavior of the new sub-pixel location algorithms is influenced by the size of Gaussian window, an optimal size of calculating window is given out. In order to reduce the negative effects of noise, a weighted scheme for centroiding is considered, and it improves the accuracy of extracting the star position on the focal plane array. The simulation results show that the accuracy of the star position can reach 1/150 pixel, and the optimal size of window for calculating is 5×5, which is under the 10% star magnitude noise. Star trackers determine the attitude of the spacecraft from some reference stars in the celestial sphere. The process of determination attitude can be divided into four steps. Firstly, the star tracker takes the star photo using the photo detectors (CCD or CMOS) in an arbitrary direction. Secondly, the microprocessor operates the image processing in the field of view (FOV) and uses the location algorithm to calculate the stars' centroid coordinates. Thirdly, star coordinates are put into the star pattern recognition algorithm and star constellations are matched in the FOV to a star catalogue covering the entire firmament. Finally, this information is adopted to calculate the attitude matrix and attitude angle and rate are delivered to estimate to the attitude control system. The second part of the process is a most crucial step, because the errors in centroid estimation dictate the overall pointing accuracy of the star tracker. A number of methods are available for estimating the centroid locations. The center of mass (COM) and some modifier algorithms based on the COM are the most commonly used techniques owning to its simplicity and robustness(3). But these algorithms do not consider the star spread numerical model and have relative low accuracy. A new Gaussian Fitting algorithm is proposed in this research. It takes into account the point spread function model and while computationally more intensive is more accuracy than the center of mass. Moreover, we use a weighted scheme to reduce the negative effects of noise, and give out the optimal size of window for calculating under the 10% star magnitude noise.