Research On Star Extraction And Star Pattern Recognition Technology In Stray Light Background

Star sensor is the core component of high precision autonomous attitude measurement for aerospace vehicles.The attitude data information provided by star sensor is generally in angular second magnitude.It is the absolute attitude measurement component of spacecraft with the highest accuracy,the most precise and the smallest drift so far.It has been widely used in earth observation,deep space exploration and other fields,and has produced significant economic and military benefits.In the future,it will have more extensive application prospects.With the improvement of aerospace vehicle's requirement for attitude determination and target tracking,it is an inevitable trend to develop star sensors with higher performance,which has become the focus of attention and research for talents in aerospace field at home and abroad.Star pattern recognition is the core step for star sensor to realize high precision autonomous attitude measurement of aerospace vehicle.Recognition speed and recognition accuracy are the key parameters for evaluating the performance of star sensor.Therefore,the study of star pattern recognition technology is helpful to develop star sensor with higher performance.However,the research in related fields in China is still relatively backward.There is a big gap between star extraction method and star pattern recognition algorithm and the international leading level.In the face of complex background(including stray light,noise,etc.),star pattern recognition is often not successful.In order to realize star pattern recognition in cluttered light background,this paper focuses on three key technologies: star pattern simulation,star point extraction and star pattern recognition in cluttered light background.The main research work is as follows:In the aspect of star pattern simulation under stray light background,the source and characteristics of stray light are analyzed in detail according to the star pattern with stray light background obtained from experiments.The simulation formulas of stray light which obey three different distribution models are given,including uniform distribution,linear distribution and Gauss distribution.Based on the traditional star pattern simulation technology,a star pattern with stray light background is proposed.Three star patterns with clutter are simulated based on MATLAB,which have very high similarity with the actual star pattern.In the aspect of star extraction under stray light background,the most classical star extraction method,namely global adaptive threshold method,is introduced.Using this method to extract star points from star images with stray light,the accuracy of the extraction results is poor.Then,based on digital image processing technology,a simple and feasible iterative method of gray transformation is proposed,which effectively eliminates the background of stray light and achieves star extraction.The simulation results show that the principle of the star extraction algorithm proposed in this paper is simple and easy to implement.Although it takes longer than the traditional method,its accuracy is higher than the traditional algorithm,and it has certain practicability.In star pattern recognition,the existing star pattern recognition algorithms are summarized and evaluated.Based on the traditional triangle algorithm,an improved triangle algorithm is proposed.The space stereo angle method is used to divide the celestial sphere and realize the uniformity of navigation star library.An optimized selection method of observation triangle is proposed,which improves the constraints of the original method,simplifies the program flow and improves the stability and accuracy of attitude calculation by using the recognition results.At the same time,an improved triangle matching algorithm is proposed,which is compared with the original method.The redundant matching checking process is added to improve the stability and matching efficiency of star pattern recognition algorithm.The simulation results show that the recognition accuracy and speed of this algorithm are still higher than those of the traditional algorithm under three different conditions: magnitude error,star position error and "missing star".The stability and accuracy of attitude calculation using the recognition results are better than those of the traditional method and other improved algorithms.