Abstract:

When the star sensor rotates with high dynamics, the weak energy of the star is spread on more pixels, the low signal-to-noise ratio (SNR) makes detection be difficult from single frame images and degrades the accuracy of the centroiding algorithm. A series of techniques intended to overcome this limitation is proposed. The first technique is to predict the position of stars and estimate the sub-window around a star. The second technique is to obtain the accurate velocity that the star image slides along the pixel array by a block matching algorithm with 6~8 stars in contiguity frames. The connectivity analysis of the pixels only deals with a small part of the image pixels in sub-window so as to ensure real-time ability. The third technique is to establish a motion compensation model. When stars sweep the pixels that the number of pixels is obtained by the first technique, the every value of grey scale of pixels expresses by several adjacent pixels instead of by itself, and the energy accumulation process is completed along the motion track of stars.