摘要
提出了一种基于QR分解的广义辨别分析算法,并将其用于雷达目标一维距离像识别.与传统用奇异值分解获取目标特征子空间的方式不同,新算法运用核修正格兰-施密特正交化过程直接提取最优投影变换矩阵,不仅有效地地保留了类内散度矩阵最具辨别力的零空间信息,同时使所求解在数值上更稳定.对3种实测飞机数据的分类结果表明,所提方法不仅在识别性能上优于传统方法,而且在一定程度上降低了算法的计算复杂度,提高了系统的实时性能.
A new generalized discriminant analysis （GDA） method based on QR decomposition was proposed, which would be used in radar target recognition with one-dimensional range profile. Different from the traditional approach of solving GDA by singular value decomposition （SVD）, the new algorithm utilizes kernel modified Gram-Schmidt （KMGS） orthogo- nalization algorithm to extract the optimal transformation matrix directly, which can not only effectively hold the most dis- criminant information in the null space of within-class scatter matrix, but also make the solution more stable in numeric. Experiments on three measured airplalns data show that the proposed method achieves better recognition performance than traditional GDA, while it has lower costs in computation partly, thereby, the real-time performance is improved.
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
2007年第3期205-208,共4页
Journal of Infrared and Millimeter Waves
基金
国家自然科学基金资助项目(60372022)
新世纪优秀人才支持计划资助项目(NCET-05-0806)
关键词
雷达目标识别
广义辨别分析
核修正格兰-施密特正交化
特征提取
一维距离像
radar target recognition
generalized discriminant analysis
kernel modified Gram-Schmidt orthogonalization
feature extraction
one-dimensional range profile
作者简介
刘华林（1977-），男，湖南湘潭人，博士研究生，主要研究方向为雷达目标识别、信号处理、机器学习．
