High Accuracy Star Camera Image Processing Technology

Star camera determines the current attitude of itself and its carrier by detectingdifferent locations but precisely known stars on the celestial sphere. Fromperspective of estimation principle, star camera belongs to the generalized starsensors. It possesses all advantages of the star sensor and is currently the mostaccurate attitude estimation system. Research on high accuracy star camera is ofgreat importance in the precise positioning of mapping and reconnaissance satellites.To improve the accuracy of attitude estimation, in terms of software, a variety ofimage processing method are applied to obtain a more accurate star centroid locationand in terms of hardware, multi-field star camera is adopted.In this paper, the principle of star camera is presented and its technicalparameters such as field of view, detection probability, detection sensitivity andestimation accuracy are analyzed. At the same time, these parameters are set afterdiscussing their relationship. Influence factors of estimation accuracy are analyzedand it is proved that the most fundamental method for its improvement is to increaseestimation accuracy of centroid location.Several major factors that affect the star centroid location estimation accuracyare system error of its algorithm, noise of image sensor and star tailing smear causedby relative motion. Relationships between the systematic error of centroid locationand Gaussian radius, the real position of star point, sampling window are simulated which the Gaussian radius and the size of sampling window are determined.Centroid error model caused by image sensor noise and dynamic star point modelare presented. Meanwhile, star image under noise and dynamic star image areobtained by simulation, influences of noise and relative motion on centroid locationaccuracy are analyzed.According to the characteristics of the star image; for relatively static imagewith high SNR, centroid star points are extracted directly by Gaussian low-passfiltering, adaptive threshold segmentation, cross-projection method and thresholdcentroid method; for relatively static star map with low SNR, star centroid locationsaccuracy are improved by multi-frame mean to reduce noise; for dynamic star imgewith high angular rate, its spectrum is analyzed to obtain motion parameters andthen centroid compensation is made; for star imge with low angular rate, triangleimage matching method is applied to estimate the motion parameters for centroidmotion compensation.Multi-FOV star camera which can achieve high precision attitude estimationon all three axes is studied. Vectors of navigation stars in multi-FOV can betransformed into the same field through coordinate transformation method, and thenattitude of satellites can be achieved using QUEST algorithm.