基于协同编码分类器的SAR目标识别方法

SAR Target Recognition Based on Collaborative Representation-based Classification

合成孔径雷达(Synthetic Aperture Radar,SAR)自动目标识别(Automatic Target Recognition,ATR)是SAR图像解译的关键技术之一。本文提出了基于协同编码分类器的SAR目标识别方法。协同编码采用所有类别训练样本构建的全局字典最优重构测试样本,并根据各类别的重构误差判定目标类别。相比稀疏表示的方法,协同编码的策略可以提升少量训练样本对于测试样本的表示能力。针对SAR目标识别,训练样本的资源十分有限。因此,协同编码表示更为适用。采用MSTAR十类目标数据集在多种条件下进行了目标识别实验并与其它分类器进行了对比。实验结果表明,本文方法在标准操作条件、型号变化、俯仰角变化以及少量训练样本等条件下均可以觉得优越的识别性能,证明了其有效性。

Synthetic aperture radar(SAR)automatic target recognition(ATR)is one of the key techniques in SAR image interpretation.This paper proposes a SAR target recognition method based on collaborative representation-based classification(CRC).Collaborative representation reconstructs the test sample using the global dictionary combined by the training samples of all the classes and decides the target type based on the reconstruction errors of individual classes.Compared with spares representation,collaborative representation can improve the representation capability of small training set.As for SAR target recognition,the training source is hard to access.Therefore,CRC is more suitable for SAR target recognition.Experiments are conducted on public MSTAR(Moving and Stationary Target Acquisition and Recognition)dataset with 10 classes of targets.According to the experimental results,the proposed method could achieve superior performance under the standard operating condition,configuration variance,depression angle variance,and reduced training set,which validates its effectiveness.