| With the improvement of the resolution of satellite images, the position accuracy of the images is required to increase correspondingly to give full scope to the advantages of high-resolution. At present, directly acquiring the accurate position information based on the orbital position and attitude information of the satellite but not the ground control points becomes an important trend in the application of the remote sensing satellites at home and abroad. However, high directly positioning accuracy inevitable requires accurate attitude information. Many attitude sensors such as sun sensor, infrared earth sensors, gyroscopes and so on, which were used for the spacecraft, are unable to satisfy the demand of accuracy, so the star sensor with highest accuracy is more and more popular for the attitude determination of the remote sensing satellite. At present, most foreign high-resolution remote sensing satellites such as IKONOS, QuickBird, etc. use the star sensor as one of the attitude determination devices, and the direct positioning accuracy of these satellites can reach 10 meters or better. Currently, a few domestic high-resolution remote sensing satellites also use the star sensor to determine attitude, but the direct positioning accuracy is still fairly far behind. One of the important factors is the accuracy and reliability of attitude determination based on star sensor is below the international most advanced level.The accuracy and the reliability of the attitude determination based on star sensor not only depend on the hardware of the star sensor, but also have an important relationship with the software of the star sensor. Since the star sensor was developed in the 1940s, many foreign researchers begin to do a great deal of fruitful works on the related algorithms. Especially when the hardware of the star sensor was updated, a great number of novel algorithms emerged to guarantee the high precision and reliability of the star sensor. In the 1980’s more and more attention is paid to the star sensor at home, and some research institutions such as the Beijing Astronomical Observatory, ChangChun Institute of Optics Fine Mechanics and Physics, ChengDu Institute of Optics and Electronics, China Aerospace Corporation Organization, National Defense University, Harbin Industry University and so on, have done a series of studies on the design of the algorithms. But compared to the international advanced level, the domestic algorithm researches are still fairly far behind. The factors affecting the attitude accuracy and reliability of star sensor are analyzed firstly in this dissertation, and then the critical technology such as the selection of the navigation stars, the star centroiding, the star identification, the calibration of the stellar camera, attitude computation and so on are analyzed in depth. In order to improve the attitude accuracy and reliability, many novel algorithms are proposed based on these analyses of the available algorithms, and a set program for attitude determination is worked out. A ideal result is obtained at last. The main contribution and innovation are as follows:1. According to the principle of attitude determination based on the star sensor, the error sources affecting the attitude accuracy and reliability are analyzed. The regularity of these factors affecting the attitude accuracy is analyzed by the simulation experiments, which can provide the theoretical reference for the subsequent researches of the critical technology.2. The critical technology of the pre-processing of star identification are summarized and analyzed, based on which, some improvement and innovation are proposed.1) A navigation star selection algorithm based on the angular gridThe selection algorithms of the navigation stars are summarized and analyzed, based on which, a simple and feasible selection algorithm—a navigation star selection algorithm based on the angular grid is proposed, based on which, the stars with uniform distribution can be extracted rapidly as the navigation stars.2) A novel flow of star centroidingAccording to the characteristics that the noise of star image acquired on the ground is large, a novel flow of stars centroiding is proposed. The flow can integrates the edge and gray information to get the centroid of star image point effectively.3. On the basis of the summary of the star identification algorithms, two on-orbit calibration algorithms of the stellar camera are improved.1) A central star identification algorithm based on the matching probabilityA central star identification algorithm based on the matching probability is proposed. This algorithm only uses the angular distance as the matching element. It constructs radicalized star pattern and identify the central star by the statistics of appearance times of star Index. When the error of the angular distance is low, the successful identification percentage is very high.2) A star identification algorithm based on the directed circularityWhen the error of the angular distance is large due to the low calibration accuracy, a star identification algorithm based on the directed circularity is put forward. It constructs the multi-dimensional matching pattern—the directed circularity, and set the matching threshold to assure that the right items are not rejected, and the longest matching chains is adopted to get the correct result.4. The stellar camera calibration algorithms are summarized and analyzed, based on which, the two original algorithms are improved.1) The on-orbit calibration of the stellar camera considering the distribution of star image pointsWhen the space resection method based on multi-images is employed for camera calibration, the distribution of "the control points" has impact on the accuracy of attitude. The area percentage of convex hull out of the full image is proposed to select the star image with good distribution. The simulation experiment indicates that it also can further improve the calibration accuracy.2) An improved on-orbit calibration method based on observational vectorAs the on-orbit calibration method based on observational vector only calibrates the elements of the inner orientation, the distortion of the lens is considered in this dissertation. The formulas including the distortions are deduced. The experiment proves that the calibration accuracy can be improved when the Len has the distortion.5. An improved attitude determination algorithmThe attitude determination algorithms are summarized, and the method which uses the linear equation to solve the elements of rotation matrix is introduced, based on which, an improved attitude determination algorithm is proposed. It linearizes the collinearity equations, and the attitude can be solved by the iterative process. The experiment indicates the attitude accuracy is improved.6. Ground experiment verification and accuracy estimationThe stellar images acquired on the ground are used to verify the proposed algorithms above and a predictable Kalman filtering algorithm. The rotation of the Earth is applied to estimate the attitude accuracy. The Experiment indicates the attitude accuracy is higher than the current domestic highest accuracy the literature can consult.From the above, in order to improve the reliability and accuracy of attitude determined by star sensor, the critical technology affecting them can be analyzed and studied, and the content of the researches covers the navigation stars selection, star centroiding, star pattern recognition, the on-orbit calibration of the stellar camera, the attitude determination and so on. Based on the summary of inspired ideas in the available researches, combined with specific research problems, a few novel or improved algorithms are proposed, which provide the solid foundation for the ideal experiment result in this dissertation. |
PDF Full Text Request