一种改进的三角形星图识别算法  

A Modified Triangle Star Pattern Recognition Algorithm

作　　者：伍玲玲[1] 杨静[1,2] 

机构地区：[1]北京航空航天大学自动化科学与电气工程学院,北京100191 [2]北京航空航天大学控制一体化技术国家级重点研究实验室,北京100191

出　　处：《宇航学报》2011年第8期1740-1745,共6页Journal of Astronautics

基　　金：国家自然科学基金(60704006)

摘　　要：传统的三角形星图识别算法冗余匹配多,对噪声鲁棒性差,为此,在传统三角形算法的基础上,提出了一种改进的算法。该算法思想如下:将观测三角形三个角距组成的平面的法向量向最优主轴投影,利用投影值进行索引,得到候选导航三角形后,利用角距进行匹配,若存在冗余匹配,再利用归一化的星等消除冗余匹配。实验结果表明:与传统的三角形算法相比,两者识别所需时间基本相同,但是改进的算法对噪声的鲁棒性更好,冗余匹配更少,识别率更高。改进的算法在质心偏差为0.5像素,星等噪声为0.5Mv时,识别率高达97.7%,传统三角形算法在质心偏差为0.5像素时,识别率仅为91.4%。The traditional triangle star identification algorithm has a higher redundancy and a poor robustness to noise,a modified triangle algorithm based on traditional one is proposed.By the proposed algorithm,a planar normal vector formed by three angles of the observed triangle is projected to the principal axis to get a projection value.Some navigation triangles within some scope are selected by using the projection value,and the three angles between the observed triangle and the navigation triangles are compared,if exists redundancy,a normalized magnitude is used to eliminate it.Compared to the traditional triangle approach,the modified algorithm has a couple of advantages,including higher rate of correct star pattern recognition,lower redundancy and better robustness.The modified algorithm has the identification rate of 97.7% when the position error is 0.5 pixels and the noise of star magnitude is 0.5Mv,while the traditional algorithm is 91.4% when the position error is 0.5 pixels.

关 键 词：星图识别 平面法向量 冗余匹配 归一化星等 

分 类 号：V448.22[航空宇航科学与技术—飞行器设计]