Research Of Star Identification Algorithm Based On Star Sensor

The attitude sensor can measure and control the attitude of the aircraft.Compared with the general attitude sensor,the star sensor has the characteristics of high precision and autonomous navigation.At present,it has become the preferred attitude measurement instrument for space vehicles.Star identification is a key technology for attitude measurement and determination of star sensors,so the research on star identification is important and meaningful.The recognition algorithm for star pattern mainly includes the process of star library,image processing and image matching.The paper explores ways to improve the robustness of the Star identification algorithm and reduce the capacity of the star library.The paper completes the whole process of image simulation and processing,image matching and algorithm evaluation.The SAO star list was selected as the basic star chart,and the threshold was set to 6.5 after that the binaries were merged into one star.Using these data to simulate the star map,the simulated image was added with noise,and then the filtering algorithm and threshold algorithm were used for image processing.Finally,centroid extraction algorithm is selected,then the centroid method,the square-weighted centroid method and the threshold centroid method are analyzed,and three algorithms are evaluated.The traditional triangle algorithm and grid algorithm are introduced.Based on the grid algorithm,this paper presents a new method that is based on the star identification algorithm of the BP neural network.The row and column numbers of the stars in the grid are recorded and the number of no stars is calculated.This forms eigenvectors as an input sample,which is identified after training using the BP network.On the basis of traditional triangles,using the star's brightness and distance information,this paper selects four companion stars and uses angular distance information to complete star identify.Finally,algorithm evaluation of star identification is mainly divided into robustness,recognition time and storage capacity.After simulation experiments,the star identification algorithm based on neural network has better robustness.The recognition speed has been improved and the storage capacity has been reduced.However,more time is needed to train it.Besides the improved triangular algorithm improves the recognition speed and reduces the capacity of the star library.The experimental results show that the star identification algorithm based on neural network and the improved triangle algorithm have better robustness and reduce the star library capacity.And the improved triangle algorithm has better applicability.