联合多特征的MSTAR数据集SAR目标识别方法  

SAR target recognition of MSTAR dataset based on joint use of multiple features

作　　者：赵高丽[1,2] 宋军平 ZHAO Gaoli;SONG Junping(School of Information Engineering,Henan Institute of Science and Technology,Xinxiang 453003,China;School of Information Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]河南科技学院信息工程学院,河南新乡453000 [2]武汉理工大学信息工程学院,湖北武汉453000

出　　处：《武汉大学学报（工学版）》2022年第7期732-739,共8页Engineering Journal of Wuhan University

基　　金：河南省教育厅资助项目(编号:2019SJGLX124)。

摘　　要：提出联合Zernike矩、核主成分分析(kernel principal component analysis,KPCA)及单演信号特征的合成孔径雷达(synthetic aperture radar,SAR)目标识别方法,3类特征分别描述SAR目标几何形状特征、投影特征及图像分解特征,联合使用可以有效增强对目标的描述。分类阶段采用联合稀疏表示对测试样本的3类特征进行表征。对于3类特征的输出误差矢量,采用多组权值矢量进行加权融合,进而根据最终融合的误差判定测试样本的目标类别。基于MSTAR(moving and stationary target acquisition and recognition)数据集的标准操作条件、扩展操作条件及少量训练样本条件对所提方法进行测试,结果验证了所提方法的有效性、稳健性及对少量训练样本的适应性。A synthetic aperture radar(SAR)target recognition method is proposed via joint use of Zernike moment,kernel principal component analysis(KPCA),and monogenic signal.The three features describe the geometrical,projection,and image decomposition properties of SAR,and their joint use can enhance the depictions of the targets.In the classification phase,the joint sparse representation is employed to represent the three features of the test samples.For the output error vectors of the three types of features,multiple weight vectors are used to fuse them.The target label of the test sample is finally determined based on the fused reconstruction errors.The proposed method is tested based on the standard operating condition,extended operating conditions,and few training samples of moving and stationary target acquisition and recognition(MSTAR)data set.The experimental results confirm the effectiveness,robustness and adaptivity to few training samples of the proposed method.

关 键 词：合成孔径雷达 目标识别 ZERNIKE矩 核主成分分析 单演信号 联合稀疏表示 MSTAR数据集 

分 类 号：TN957[电子电信—信号与信息处理]