基于多特征匹配的快速星图识别

Fast star identification algorithm based on multi-feature matching

随着星敏感器探测灵敏度的提高,导航星表中恒星的数量急剧增加,导致星图识别的识别速度和识别率降低。因此,为了提高星图识别的识别速度和识别率,文中在三角形算法的基础上提出了一种基于多特征匹配的快速星图识别算法。首先,采用天球的内接正二十面体法对预处理后的星表进行分区。然后,将特征三角形的边长以及外接圆和内切圆半径的乘积作为特征值构建导航特征库,并根据后者的哈希函数对特征库进行分块。在识别的过程中,利用观测三角形外接圆和内切圆半径乘积的哈希函数实现导航特征库子块的快速定位,并在该子块内采用多特征匹配的方法得到观测三角形的识别结果,最后根据该结果确定星敏感器视场所包含的天球子区域,并在子区域内完成视场中其它导航星的识别。实验结果表明,文中算法的识别性能与导航特征库的分块数有关,在选择合适的分块数后,与常用三角形算法相比,算法在识别速度,识别率以及对星等误差和假目标的鲁棒性等方面具有明显的优势,算法的平均识别时间和识别率分别为17.161 ms和98.58%,满足星敏感器对高识别速度和识别率的要求。

With the improving detection sensitivity of star sensors,the number of stars in star catalogs has increased dramatically,reducing the speed and rate of star identification.To improve the identification speed and rate,a fast star identification algorithm based on multi-feature matching and built upon the foundation of the triangle algorithm was proposed.First,the preprocessed star catalog was partitioned using the inserted icosahedron of the celestial sphere.Then,the navigation feature library was constructed with the sides and the product of the radii of circumcircles and incircles as feature values.In addition,the navigation feature library was stored in blocks according to the hash function of the latter feature value.In the process of star identification,the product of the radii of the circumcircle and incircle in the observation triangle was used to rapidly locate the block of the navigation feature library,and then the observation triangle was identified in the block by using multifeature matching.Finally,the sub-regions of the celestial sphere in the field of view were obtained,and then other navigation stars were identified in the sub-regions.The experimental results indicate that the identification performance of the proposed algorithm is related to the number of blocks.Based on a reasonable number of blocks,the proposed algorithm has advantages in identification speed and rate as well as in its robustness to star magnitude noise and false stars,compared with common triangle algorithms.The average identification time and rate of the proposed algorithm are 17.161 ms and 98.58%,respectively,which can meet the star sensor requirements for high identification speed and rate.