基于多信息字典学习及稀疏表示的SAR目标识别

发布时间：2018-04-05 15:18

本文选题：合成孔径雷达　切入点：目标识别　出处：《系统工程与电子技术》2015年06期

【摘要】：为了提高合成孔径雷达(synthetic aperture radar,SAR)图像中目标变体的识别性能,在鉴别字典学习及联合动态稀疏表示模型的基础上,提出一种基于多信息字典学习及稀疏表示的SAR目标识别方法。在训练阶段,采用鉴别字典学习LC-KSVD方法分别对目标图像域幅度信息及目标频域幅度信息进行字典学习。在测试阶段,结合训练阶段学到的2种信息的字典及测试目标的2种信息,采用联合动态稀疏表示模型求解2种信息下的稀疏表示系数。最后,根据2种信息下的重构误差实现对测试目标的识别。使用MSTAR数据集对算法进行验证,结果表明,新方法相对于现有的方法能够达到更好的识别性能。
[Abstract]:In order to improve the recognition performance of target variants in synthetic aperture radar (SAR) images, a method of SAR target recognition based on multi-information dictionary learning and sparse representation is proposed on the basis of discriminant dictionary learning and joint dynamic sparse representation model.In the training stage, the discriminant dictionary learning (LC-KSVD) method is used to study the target image domain amplitude information and the target frequency domain amplitude information respectively.In the test stage, combining the dictionary of the two kinds of information and the two kinds of information of the test object, the sparse representation coefficient under the two kinds of information is solved by using the joint dynamic sparse representation model.Finally, according to the reconstruction error of two kinds of information, the recognition of test target is realized.The MSTAR dataset is used to validate the algorithm. The results show that the new method can achieve better recognition performance than the existing methods.
【作者单位】：西安电子科技大学雷达信号处理国家重点实验室;
【基金】：国家自然科学基金(61201292,61322103,61372132) 全国优秀博士学位论文作者专项资金(FANEDD-201156) 国防预研基金中央高校基本科研业务费专项资金资助课题
【分类号】：TN957.52

【参考文献】