摘要
提出了以凸多边形为基元、完全不依赖于星等的星图识别算法。首先,构造全部导航星的 14°视场模式,共9 176个。然后逐一把这些视场中的导航星投影到像平面,得到一系列的点。依其坐标排序,采用由平面上的点生成凸多边形的算法,就能得到唯一的、以恒星为顶点的凸多边形。在导航星表中以凸多边形为储存单元,其内容为凸多边形的边和顶角。为此,设计了以导航星凸多边形和观测凸多边形之间距离最小为准则的识别算法。针对星光制导的导弹观星时间很短,提出了根据弹道生成弹上导航星表的方法,其导航星表只需存储30颗星的 75 个凸多边形。仿真结果表明:在任意视场中,基于凸多边形的星图识别成功率高于99%,并具有较强的鲁棒性。
Star pattern identification algorithm based on convex polygon with star magnitudes-independent is presented for strap-down celestial guidance. First, all stars within the field of view (FOV) of 14° are found and the number of all FOV is 9176 obtained by going through the catalog of the navigation stars. Subsequently, the stars within the FOV are projected into the image plane of the star sensor; by sorting according to coordinates of the projected points and using a technique to generate the minimum bounding convex polygon, a unique convex polygon with stars as its vertexes can be obtained and used as a star identification pattern, which is stored with line segments and interior angles that are marked beginning from the shortest segment along clockwise direction. As the star sensor of ballistic missile works only for short time, the navigation stars in the missile can be reduced to 30, and the convex polygons can be reduced to 75 for different trajectories. Finally, numeral simulation shows that the star pattern identification algorithm based on the convex polygon has high robustness and the ratio of identification is above 99%.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2005年第1期88-94,共7页
Optics and Precision Engineering
基金
第二炮兵工程学院预研课题资助
关键词
星敏感器
星图识别
凸多边形
算法
弹上星表
Algorithms
Computer simulation
Missiles
Numerical methods
Pattern recognition
Robustness (control systems)
Sensors
