Analysis And Research On The All-sky Autonomous Fast Star Matching Methods

As a type of angular sensors with high precision, star sensor has advantages of high autonomy and no accumulative error, and can provide accurate azimuth and reference for spacecraft, thus playing an important role in aerospace industry and other fields. As star sensor‘s maximum detection range of stellar magnitude raises, the number of stars in guide star catalog increases exponentially and the performance of star identification algorithm decreases quickily. In order to realize a fast and accurate all-sky identification of star sensor, it is necessary to give a deep analysis in the principle of star matching methods. The thesis is based on the Joint Laboratory of Ocean-based flight vehicle Measurement And Control‘s open fund project ?Analysis and research on the all-sky autonomous fast star matching methods‘, which proceed from the basic principle of star identification algorithm, two typical all-sky autonomous star identification algorithms are analyzed theoretically and verified by simulation experiments, and a more robust star identification algorithm are explored on that basis. The main works are listed as follows:(1) The basic principle of star identification algorithm is introduced and comparatively implementation steps are elaborated.(2) In order to emphatically analyses the factors which effect the performance of star identification algorithms, two typical algorithms- triangle algorithm and grid algorithm are simulated by using MATLAB, the influence of position error, magnitude error, lost stars and fake stars on identification rate of the two algorithms is studied. The simulation results show that the triangle algorithm is easy to realized, but it has redundant match, and the identification time is slow; grid algorithm has a fast identification time and the capacity is small, which has obvious superiority.(3) According to the simulation results, a high limit stellar magnitude star identification is realized by using grid algorithm. In the condition of 9 stellar magnitudes, the star identification is realized by using MATLAB, the average identification time is 1.342 seconds, the identification rate is nearly 100%, which can meet the project‘s demands.(4) In order to solve the problem of rotating the picture with a chosen neighbor-star when creating the star pattern by using grid algorithm, a rotation- invariant based star identification algorithm is proposed. The algorithm uses the idea of pattern-recognition based star identification algorithm, builds a rotation independence pattern for every star, and uses the improved Hausdorff distance for the matching. The simulation results show that the rotation- invariant based star identification algorithm has a better robustness in position errors and interferential stars than grid algorithm, and the algorithm has good recognition effect.