摘要
针对高分辨距离像(HRRP)可分性低和维数高的问题,提出一种新的雷达自动目标识别(RATR)方法:dLDA&SVM。先采用直接线性判别分析在HRRP的幅度谱空间进行特征提取,然后在子空间中采用角域均值模板库训练one-against-all支撑向量机(SVM)多类分类器进行目标识别。并设计了最短距离分类器与SVM分类器比较。基于外场实测数据的实验结果表明,与LDA幅度谱子空间法,幅度谱原空间法相比,dLDA&SVM可显著降低数据维数并提高识别性能。
High resolution range profile （HRRP） has the problems of low separability and high dimensionality. A novel radar automatic target recognition （RATR） method, i. e. , dLDA＆SVM, is presented. Firstly, a direct linear discriminant analysis （dLDA） is used to perform feature extraction in the amplitude spectrum space of HRRP, and then the mean of each azimuth in the resulting amplitude spectrum subspaee is used to train an one-against-all support vector machine （SVM） rnulti-class classifier for target recognition. A shortest distance classifier is also designed for comparing with the SVM classifier. Experimental results for measured data show that comparing with the target recognitions in the amplitude spectrum subspace of LDA and the original amplitude spectrum space, dLDA＆SVM can remarkably reduce data dimensionality and improves recognition performance.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2008年第10期1815-1818,共4页
Systems Engineering and Electronics
基金
中意科技合作项目基金资助课题
关键词
雷达自动目标识别
特征提取
直接线性判别分析
高分辨距离像
支持向量机
radar automatic target recognition （RATR）
feature extraction
direct linear discriminant analysis
high resolution range profile （HRRP）
support vector machine （SVM）
