基于核主分量分析的高分辨雷达目标特征提取与识别  

Target feature extraction and recognition for high-range recognition radar based on the KPCA method

作　　者：丛瑜[1] 肖怀铁[2] 付强[2] 

机构地区：[1]海军航空工程学院,山东烟台264001 [2]国防科技大学电子科学与工程学院ATR实验室,长沙410073

出　　处：《电光与控制》2008年第2期31-35,38,共6页Electronics Optics & Control

基　　金：国家自然科学基金资助(60572138);“九七三”项目(5131402)

摘　　要：研究了核主分量分析(KPCA,Kernel Principal Component Analysis)在高分辨雷达目标特征提取与识别中的应用。首先讨论了KPCA算法原理,然后将KPCA应用于雷达目标距离像特征提取,并采用支持向量机进行分类,提出了基于核主分量分析的高分辨雷达目标特征提取与识别方法。在核函数的选取上构造了一个组合核函数,最后用4类目标数据进行了实验,并与采用高斯核函数方法进行了比较,实验结果表明,该方法能够提高目标识别性能。We studied the application of Kernel Principal Component Analysis （KPCA） in target＇ s feature extraction and recognition of High Range Resolution Profile（HRRP） radar. The principle of the KPCA algorithm is discussed at first. Then the KPCA method is used in target extraction for high range recognition radar. Support Vector Machine（SVM） is used for sorting. Then the method for target feature extraction and recognition of HRRP radar is put forward based on KPCA. A co-kernel function is constructed. Experiments were carried out on four types of targets. The method was compared with the Gaussian kernel method. The result showed that the method put forward can improve target recognition performance.

关 键 词：核主分量分析 雷达目标识别 一维距离像 支持矢量机 

分 类 号：V249.3[航空宇航科学与技术—飞行器设计]