Researh On The Key Techniques Of High Precise Double-FOV Star Sensor

Star sensor is a kind of high precise attitude determination equipment for spacecraft. In the application of the spacecraft autonomous navigation, the size of the field of view (FOV) of the star sensor is conflicted between the precision and the detection ability of the star sensor. To improve the precision of the star sensor and enhance its detection ability, a double-FOV star sensor is proposed in the architecture in this dissertation, which is comprised of two image sensors and two fields of view in a single star sensor. Based on the star sensor, the research is carried out in algorithms and experiments, on fast star identification algorithms and high precise attitude estimation algorithms. The research is listed as follows in detail:The principle of the star sensor is analyzed and a new double-FOV star sensor which is comprised of two image sensors and two fields of view is proposed in the architecture and in the principle solution. The algorithms in how to use the star images taken in the two fields of view in the processes of image collection, star acquisition and centroiding, star identification and attitude computing are discussed. Then the detection ability of the star sensor and the factors that affect the detection ability of the star senor is analyzed, and the methods to improve the detection range of the star sensor are discussed. By analyzing the features of the targets and the characters of the responding curve of the star sensor, the method to improve the detection ability of the star sensor by adaptively adjusting the integrate time is proposed. In the method, the statistical grey value of the brightest fixed star is used as the reference of the image brightness, and the linear section of the responding curve of the star sensor is used to compute the expected integrate time. The experiments show that the star sensor with adaptive integrate time can take good photos on the objects with varying background, and the detection ability of the star sensor is enhanced.The usually used star patterns in the star identification are analyzed. As the angular distances and the angles between the lines of the stars can be used to denote the polygon pattern, a new star-shape pattern is proposed, which is composed of a reference star and its neighboring stars centered it. The pattern is proven to be equivalent to the polygon pattern in geometry, and can present the distribution of the stars in the star image directly. An angular polar coordinate system is proposed to present the star-shape pattern easily, in which the coordinates are denoted by angular values, and the star-shape pattern is invariant to rotation and translation. Using the principles of the grid algorithm, a grid algorithm is proposed based on the angular polar coordinate system. Then a new angular polar coordinate system based grid algorithm for star identification is proposed by using the grid algorithm in star-shape identification in the angular coordinate system. The simulations demonstrate that the algorithm in the angular polar coordinate system is superior to both of the original grid algorithm and the triangle algorithm in the rectangle Cartesian coordinate system in rate of success, identification speed and robustness to the measuring noises.The quaternion based attitude estimation algorithms for star sensor are analyzed, and a modified quaternion based attitude measurement equation proposed by Choukroun et al is introduced to the attitude estimation of the star sensor, and the responding attitude estimation algorithm is proposed. The simulations show that the algorithms can yield higher precise attitude than the conventional EKF based algorithm. The attitude estimation fusion algorithms in the double-FOV star sensor are studied, and the federated kalman filtering algorithm is proven to be superior for attitude estimation fusion of the double-FOV star sensor in simulation as well as in theoretical analysis. To improve the precision of the attitude estimation in the double-FOV star sensor, the information sharing method based on the federated kalman filtering algorithm is studied. Then the matrix-form information sharing factors are proposed to use the information of the components of the attitude quaternion, and the information sharing matrix using the system state covariance and the measuring error covariance are studied. The simulations demonstrate that the attitude estimation fusion algorithms using the information sharing matrix are of high precision and high robustness.