摘要
提出了联合多层次深度特征的合成孔径雷达(SAR)目标识别方法。采用卷积神经网络(CNN)学习SAR图像的多层次深度特征。多层次的深度特征从不同方面描述原始SAR图像中的目标特性,从而为目标识别提供更充分的决策依据。为了充分发掘不同层次深度特征的独立特性以及它们之间的内在关联,采用联合稀疏表示对多层次的深度特征进行联合分类。根据各层次特征的整体重构误差判定目标类别。采用MSTAR (Moving and Stationary Target Acquisition and Recognition)公共数据集对提出方法进行了性能测试。实验结果表明,该方法的识别性能显著优于现有的SAR目标识别方法。
A synthetic aperture radar(SAR)target recognition method is proposed via joint use of multi-level deep features. The convolutional neural network(CNN) is employed to learn multi-level deep features from the original SAR image,which describe the target characteristics from different aspects. Therefore,the multi-level deep features could provide more supports for the decisions in target recognition. In order to fully exploit the individual discriminability of the multi-level deep features as well as their inner correlation,the joint sparse representation is employed as the classifier. Afterwards,the target label of the test sample is decided based on the total reconstruction errors of the multi-level deep features. The public moving and stationary target acquisition and recognition(MSTAR)dataset is used to test the performance of the proposed method. The experimental results show that the proposed method outperforms some state-of-the-art SAR target recognition methods significantly.
作者
张婷
蔡德饶
ZHANG Ting;CAI De-rao(Jiangxi Police Institute,Nanchang 330031,China;Department of Information Engineering,Shangrao Vocational and Technical College,Shangrao 334100,China)
出处
《火力与指挥控制》
CSCD
北大核心
2020年第2期135-140,共6页
Fire Control & Command Control
基金
2017年度江西省高校人文社会科学研究青年项目(JY17248)
2017年江西省高等学校教学改革研究课题(JXJG-17-77-6)。