联合多分辨表示的SAR图像目标识别方法

SAR target recognition based on joint use of multi-resolution representations

提出了联合多分辨率表示的合成孔径雷达(SAR)目标识别方法。该方法首先根据SAR图像的成像机理构造原始图像的多分辨率表示。多分辨率表示以互补的方式由粗到精地描述了目标的特性,可以为后续的目标识别提供更丰富的鉴别力信息。为了充分利用多分辨率表示中蕴含的信息,采用联合稀疏表示对其进行分类。作为一种多任务学习算法,联合稀疏表示既可以有效表示各个分辨率上的表示还可以充分发掘各个分辨率之间的内在相关性。因此,结合多分辨率表示和联合稀疏表示分类器可以有效提高SAR目标识别性能。基于MSTAR(moving and stationary target acquisition and recognition)公共数据集在多种操作条件下进行了目标识别实验,充分验证了方法的有效性。

This paper proposes a synthetic aperture radar ( SAR) target recognition method based on joint use of multi-resolution representations. The multi-resolution representations are generated according to the SAR imaging mechanism,which provide coarse to fine descriptions of the target characteristics in a complementary way. Therefore,the multi-resolution representations can provide more discriminability for the following target recognition. To fully exploit the discriminability,the joint sparse representation is employed to classify the multi-resolution representations. As a multi-task learning algorithm,the joint sparse representation not only represents each resolution properly but also exploits the inner correlations between among resolutions. Therefore,the combination of the multi-resolution representations and the joint sparse representation can effectively improve the ATR performance. To validate the effeteness of the proposed method,experiments are conducted on public moving and stationary target acquisition and recognition( MSTAR) dataset under several operating conditions