一种多视角SAR图像目标识别方法

A Multi-view SAR Target Recognition Method

针对合成孔径雷达(Synthetic Aperture Radar,SAR)图像目标识别问题,提出了一种联合多视角的方法。基于图像相关准则对多视角SAR图像进行聚类分析,获得若干视角集。分别对每个多视角子集采用多重集典型相关(Multiset Canonical Correlations Analysis,MCCA)进行特征融合,获得特征矢量。采用联合稀疏表示对各个视角集的特征矢量进行表征分类,获得决策结果。在样本丰富的MSTAR数据集上开展实验与分析,结果表明,所提方法对10类目标样本在标准操作条件、噪声干扰以及遮挡情形下均可以取得优势性能,验证了其有效性。

For synthetic aperture radar(SAR)target recognition problem,a multi-view method is proposed.For the multi-view SAR images of the same target,the clustering analysis is performed to obtain several view sets.For each view set,the multiset canonical correlations analysis(MCCA)is employed to fuse them as one single feature vector.The joint sparse representation is adopted to represent and classify the feature vectors from different view sets thus deciding the target label of the multi-view SAR images.Experiments are conducted on the MSTAR dataset with rich samples.The results confirm the validity of the proposed method for the 10-class samples under the standard operating condition,noise corruption,and occlusion.