SAR图像多层次正则化增强及在目标识别中的应用  

Multi-level SAR image enhancement based on regularization with application to target recognition

作　　者：谢晴[1] 张洪 

机构地区：[1]江苏省徐州医药高等职业学校 [2]河海大学计算机与信息学院

出　　处：《电子测量与仪器学报》2018年第9期157-162,共6页Journal of Electronic Measurement and Instrumentation

基　　金：江苏省自然科学基金(1730128)资助项目

摘　　要：正则化方法可以有效提取合成孔径雷达(SAR)图像中的目标特性,增强原始SAR图像质量。通过调整正则化系数构造多层次的正则化增强结果,并将其应用于SAR目标识别。多层次的正则化增强结果可以全面细致地反映目标散射中心的分布规律并通过互补的方式为目标识别提供更多的信息。为了充分发掘各个层次的鉴别力以及它们之间的内在联系,采用联合稀疏表示作为分类器完成目标识别。为了验证提出方法的有效性,在MSTAR数据集上进行了目标识别实验并与几类经典的SAR目标识别方法进行了对比。Regularization can effectively extract the target characteristics thus enhancing the quality of the original synthetic aperture radar( SAR) images. By adjusting the regularization parameters,multi-level regularization results can be constructed for SAR target recognition. The multi-level regularization results can reflect the distribution of the targets’ scattering center more comprehensively. In order to exploit the individual discriminability as well as the inner correlation between different levels,joint sparse representation is employed as the classifier for target recognition. To validate the effectiveness of the proposed method,experiments are conducted on public moving and stationary target acquisition and recognition( MSTAR) dataset and compared with several other SAR target recognition methods.

关 键 词：合成孔径雷达 目标识别 正则化增强 联合稀疏表示 

分 类 号：TN957.52[电子电信—信号与信息处理]