| The star sensor is an important device in navigation system.It is of great significance to study star identification,the key technology in the process of star sensor.So far,people have studied star identification for a long time,but the traditional star identification algorithm algorithms have a strong dependence on the sky region.Based on the analysis of the typical algorithm,this paper synthesizes the characteristics of the traditional algorithms,and improves the pre-processing technology of the star identification at the same time.It explores the two aspects of star point extraction and star map matching to improve the engineering practicability of the matching algorithm.In order to achieve high-precision star extraction in the initial stage of star map matching,this paper proposes a background prediction algorithm based on variable neighborhood weight.The algorithm uses the distribution characteristics of the center pixel and its adjacent pixels,and adaptively determines the best prediction weight of the algorithm through the distribution characteristics of the star and background pixels,so as to predict the star image background.After that,the centroid algorithm with threshold is used to extract star coordinates.Finally,experiments are carried out on simulated star images and actual star images.The experimental results show that,compared with other background prediction algorithms,the variable neighborhood weight prediction algorithm proposed in this paper has higher star detection rate and can extract stable and high-precision star points.At the same time,by studying the influence of different noises on the centroid position of stars,the effectiveness of the proposed algorithm is verified,which lays the foundation for the subsequent research of star image matching.In order to realize the recognition of star map in engineering application,this paper proposes a recognition algorithm based on Hausdorff distance and star topology structure.In this algorithm,the magnitude information is introduced to determine the main and auxiliary stars,and star map rough recognition is performed according to their Hausdorff distance.At the same time,a point-to-point star topology with the main star as the center and radiating to the surrounding stars is constructed.According to the characteristics of rotation angle and angular distance,the optimal matching of the star map is carried out.In view of the factors that affect the star map matching algorithm,this paper carries out the comprehensive verification of the improved algorithm on the selection of spatial range,the characteristic dimension of the navigation stars in the navigation star library,and the selection of the primary and secondary navigation stars.After a lot of tests,the optimum parameter selection range suitable for the algorithm in this paper is obtained.At the same time,different interference noises are set in this paper,and the experimental comparison between the improved algorithm and the triangle algorithm is made.The experimental results show that the improved algorithm has stronger anti-interference ability,and higher detection rate and recognition rate for stars. |
