Study On Star Centroid Extraction And Star Identification Method Based On Star Sensor

The deep space exploration makes contribution for the human to understand the origin and evolution of universe and life, to explore and expand the living space, to use the space resources reasonably, to pursue the sustainable development of the society. With the development of deep space explorations, the requirements of the autonomous navigation technology for the deep space probe go higher and higher. The existing deep space probes are highly dependent on the ground station. But the inherent defects of the ground station make it unable to maintain a long time and distance navigation performance steadily. So it needs to find a new navigation model. Celestial navigation has the incomparable advantages over the other navigation methods and can provide position and attitude for the deep space probe,which is a very important autonomous navigation model of the deep space probe.Star sensor is an absolute attitude sensor with highest measure precision in aerospace applications, which can provide accurate position and attitude for the satellite, the missile, and the spacecraft and so on. The autonomous celestial navigation technology(ACNT) based on star sensor belongs to the fully autonomous navigation model, which has more advantages, such as higher measure accuracy, no cumulative error and good invisibility. Domestic and overseas scholars have made great effort on the research of the ACNT based on star sensor in recent years. This dissertation focuses on the problems of star point extraction and star identification in star sensor. The author’s major contributions are outlined as follows:1. In order to improve the positioning accuracy and the efficiency of star centroid extraction for the stellar image, a high-precision and rapid star centroid extraction method is proposed. First, with the distribution characteristics of the star pixels, the coarse positioning of star centroid extraction is carried out according to the clustering algorithm, in which the global operation of star centroid extraction is converted into the local operation. So the scanning for the background pixels can be reduced. Then, the center coordinates of the star point achieved in the coarse positioning operation are used as the seed point. Based on the region growing algorithm, the star pixels are selected automatically. Subsequently, according to the distribution characteristics of the star point energy, two different corrective strategies are adopt to modify the grey values of the star pixels by using the plane distance between the star pixel and the seed point, in which the influence of noise on the star centroid extraction can be reduced. At last, the resolution of the star point is enhanced by using the bilinear interpolation algorithm, and the centroid coordinates of the star point can be achieved through the centroid calculation formula. So the fine positioning of star point is realized. The simulation results verify the effectiveness of the suggested method. The comparison results show that the method has higher positioning accuracy and the efficiency of the star point extraction is improved.2. With the analysis of the star identification algorithm based on Log-Polar transform(LPT), an improved autonomous star identification algorithm based on Log-Polar transform is proposed. The disadvantages of the LPT algorithm are analyzed firstly. Subsequently, the observed stars in field of view(FOV) are re-projected by using the reconstituted rectangular plane coordinate system of the stellar image, which makes it able to remain the recognition feature of the observed star when the stellar image rotates. Then the log-value of the plane coordinates of the observed star is used as the element of the feature vector, which can enhance the anti-noise performance of the recognition feature. Finally, the number of the nonzero values in the feature vector of the observed star is used to restrict the matching search scope of the feature vector in the feature library of the navigation stars. So the recognition of the observed star can be accelerated. The simulation results show that the method overcomes the disadvantages of the original algorithm. Compared with the modified grid algorithm and the LPT algorithm, the method has stronger robustness, which improves the recognition rate and accelerates the recognition of the observed star concurrently.3. Only the position information and the intensity of the star point in the stellar image can be used in star identification, and generally the intensity of the star point is considered as unstable information. Based on the plane geometry relationship between the observed star and its adjacent stars, an autonomous star identification algorithm based on one-dimensional vector pattern is proposed. The recognition feature of the observed star is established by using the position of the observed star and the angle information between the observed star and its adjacent stars. The observed stars in FOV are re-projected by using the vector direction of the observed star pattern, in which the recognition feature of the observed star remains unchanged when the stellar image rotates. During the process of the recognition feature matching for the observed star, the number of the nonzero values in the feature vector is used to shrink the matching search scope of the feature vector in the feature library of the navigation stars. The simulation results verify the effectiveness of the proposed algorithm, and it show that the recognition rate and the identification speed has been improved concurrently. Under the same conditions, the proposed method is compared with the pyramid algorithm, the modified grid algorithm and the LPT algorithm. The compared results show that the proposed method has better performance than the compared algorithms.4. In order to optimize the process of star identification, and enhance the robustness of star identification and accelerate the identification speed of the observed star concurrently, an autonomous star identification algorithm based on combined pattern is proposed. The method has combined the advantages of the radial mode and the encoding mode. With the translation and rotation invariant features of the radial mode, the recognition feature of the observed star remains unchanged. The utilization of the encoding mode makes it able to simplify the recognition problem of the observed star as the numerical comparison. According to the characteristics of the identification feature of the observed star in the combined pattern, the identification feature of the observed star in the encoding mode is used in the initial matching, in which the identification feature in the encoding mode is used to carry out the matching search. So the candidate matching results of the observed star can be achieved. Then the identification feature of the observed star in the radial mode is compared with the identification features of the candidate matching results in the radial mode. So the recognition result can be obtained. The simulation results verify the effectiveness of the proposed algorithm. Compared with the pyramid algorithm and the modified grid algorithm, the method can maintain stronger stability with the increase of the positional noise, which has improved the anti-noise performance and has higher identification speed.With the research of the dissertation, the simulation process of the stellar images is clear, which can provide the necessary experimental data. The high-precision star point extraction-rapid method, the improved autonomous star identification algorithm based on Log-Polar transform, the autonomous star identification algorithm based on one-dimensional vector pattern and the autonomous star identification algorithm based on combined pattern have been proposed. The proposed algorithms have be verified by using the simulation experiments, which have increased the positional accuracy and recognition speed of the star and improved the performance of star identification. The research results of the dissertation provide the theoretical basis for the improvement of the performance of star sensor, and can be served as an important guidance for the software design of star sensor with high accuracy and high stability.