稀疏表示系数相关性筛选多视角SAR目标识别方法  

A Multi-view SAR Target Recognition Method Based on Sparse Representation Coefficients Correlation Selection

作　　者：陈婕 CHEN Jie(Mechanical and Electrical Engineering College,Guilin Institute of Information Technology,Guilin 541004,China)

机构地区：[1]桂林信息科技学院机电工程学院,广西桂林541004

出　　处：《探测与控制学报》2025年第2期48-54,共7页Journal of Detection & Control

摘　　要：合成孔径雷达(SAR)图像处理是获取侦察信息的重要手段,当前SAR目标识别能力不高已成为制约其有效获取侦察信息的问题。针对这一问题,通过稀疏表示分类(SRC)对单一视角进行处理,获取相应的稀疏表示系数矢量。以不同视角稀疏表示系数矢量为基础,定义他们之间的相关性并构建相关性矩阵;基于非线性相关信息熵获取内在相关性最强的多视角子集;最后采用联合稀疏表示模型对选取得到的多视角进行分类,判定他们所属的目标类别。经选择得到的多视角在稀疏表示空间具有良好相关性,从而保证了联合稀疏表示分类的精度和可靠性。实验依托MSTAR数据集开展并进行分析,结果验证了所提方法的有效性。For multi-view synthetic aperture radar(SAR)image target recognition,the sparse representation-based classification(SRC)was first used to process a single view and obtained the corresponding sparse representation coefficient vector.Based on the sparse representation coefficient vectors of different views,the correlation between any two views was defined and a correlation matrix is constructed correspondingly.Then,based on nonlinear correlation information entropy(NCIE),a subset in the multi-view SAR images with the strongest intrinsic correlation was obtained.Finally,the joint sparse representation(JSR)model was used to classify the selected multiple views and determine their target category.The selected multiple views had good correlation in the sparse representation space,ensuring the accuracy and reliability of JSR classification result.The experiments conducted based on the MSTAR dataset proved the effectiveness of the proposed method.

关 键 词：合成孔径雷达 多视角 稀疏表示系数 联合稀疏表示 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]