Star Fast Extraction And High Accuracy Centroid Estimation Of Star Camera

Star Light Navigation is a kind of Celestial Navigation, it can provide the current attitude and position of the spacecraft in action by processing, analyzing and identifying the acquired star data. Lately, star sensor technology used in aircraft attitude measurement has become very popular because of its high accuracy, light weight, without attitude accumulation errors and other advantages.The key technology of modern star tracer is star location and star identification, which will directly impacts the accuracy and real-time of the navigation. Therefore, it is important to take an overall research on the technology of high-precision location and rapid recognition of star image.In this paper, the working principle and the relation between temperature and gain characteristics of the EMCCD is researched, the noise distribution and the gray characteristics of the star image are analyzed. We researched in depth the basic principles of various algorithms for star image processing and comprehensively consider the accuracy and speed. The algorithm mainly includes three steps:star image pretreatment, star identification and segmentation, star locating.Star image pretreatment mainly refers to star image denoising with specific filters, such as the adaptive filter, Gaussian filter, Wiener filter and so on. Star identification and segmentation includes extracting the star points from the background and separating each of them apart from the others. In practice, the thresholding algorithms are used to eliminate undesirable background and random noise that may decrease precision. Aiming at the deficiency of the existing thresholding methods, we described a new weighted threshold algorithm based on the estimation of optimal threshold for achieving minimal centroid error in the processed image. The impact of threshold on centroid accuracy and computing time is also presented in detail.Different from the traditional connected domain algorithm, we locate the star points with an improved cross projection algorithm. In each projection direction, the domain bound of the star points is extracted with the proposed binarization differential extremum method. In order to improve location precision, a sub-pixel locating technology is applied to obtain the position of the star points.After extracting the star points using those algorithms above, we can obtain the centroid motion curve of star image sequences. When the star targets move, the change of their gray-scale is characterized as occurring around a stationary value of a sinusoidal process. So we can get real interested objectives and its continuous tracings. The motion curve is used to estimate the flutter of star sensor so as to improve the star image quality by image restoration.