摘要
介绍了一种利用自组织特征映射(SOFM)网络的聚类功能进行全天星图识别的方法。按一定规则对全天星表的星信息进行了筛选并组建了导航星星库。从星库中提取星对角距信息和星模式信息来分别训练此星图识别系统中的两层SOFM网络,使网络具备了分类识别功能。以此系统进行的仿真识别结果表明,SOFM网络可以有效地反映星图中的复杂信息,抗噪性能明显优于传统星图识别算法;分类效果较好,能区分各种星模式;识别速度很快,约为1.5ms,可以在星图识别中发挥很好的作用。
A method that applies the clustering function of SOFM (Self-Organizing Feature Maps) network is proposed for autonomous star pattern recognition. The guide star catalog is built with the star information selected from all-sky star catalog on some guidelines. Angle distances and star model information abstracted from guide star catalog are used to train the two SOFM networks of the star pattern recognition system so that the networks have the classification and identification functions. It is concluded from the results of simulated identification that the SOFM network can reflect the complicated information among star pattern better and it appears more robust with respect to noise than conventional methods. The effect of classification is so good that it can differentiate all kinds of star models clearly and the speed of identification is so fast that it can recognize a star pattern within 1.5 ms. Therefore, it can be well applied for star pattern recognition.
出处
《光学精密工程》
EI
CAS
CSCD
2004年第3期346-351,共6页
Optics and Precision Engineering
关键词
星图识别
SOFM网络
导航星星库
神经网络
star pattern recognition
SOFM network
guide star catalog
neural network