改进金字塔算法用于小视场星图识别

 为实现小视场星图的全天自主识别，规避星矢量内积在小角距范围内区分度欠佳的问题，提出以星矢量外积作为匹配特征量对金字塔算法进行改进。分析了匹配特征量改进策略，并对改进金字塔算法涉及的基本星表预处理、匹配特征量数据库及K矢量构建、星图降噪与质心提取、星图识别流程等问题进行讨论。采用Visual C++编程实现该算法，通过仿真对其进行性能测试，并应用于小视场星敏感器。结果表明，对于小视场星图，改进金字塔算法识别成功率达96.7%，需载入内存的数据文件约26.4 MByte，识别时间平均约131.8 ms，能够满足全天自主星图识别算法的准确率高、占用资源适度、识别速度快、稳健性强等要求。

Modified pyramid algorithm for small FOV star image recognition

A modified pyramid algorithm is proposed for autonomous star identification of small field-of-view(FOV) star images in the general lost in space case, by selecting outer product of star vectors as characteristic matching quantity, in order to avoid poor discrimination of inner product. The strategy for modification to characteristic matching quantity is analyzed. And the issues related to the modified pyramid algorithm are discussed, involving preprocessing of the original star catalogue, construction of the database of characteristic matching quantity and the K vector, noise reduction and star centroiding, the star identification procedure, and so on. This algorithm is realized by Visual C++ programming. Its performance is tested by simulation. Finally, it is actually applied to star identification for a small FOV star tracker. Experimental results demonstrate that the star identification success rate of the modified pyramid algorithm is 96.7%, the size of data files loaded in memory is about 26.4 MByte, and the average time consumption of star identification is about 131.8 ms. This algorithm meets well the basic requirements of autonomous star identification in the lost in space case, such as high identification success rate, moderate resource consumption, fast identification speed, and strong robustness.