节点文献

结合最近邻和拓展稀疏表示的SAR图像目标识别方法

SAR image object recognition method combining nearest neighbor and extended sparse representation

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 徐斌曹娜王永利

【Author】 Xu Bin;Cao Na;Wang Yongli;Information Mangement Center, Nanjing Institute of Railway Technology;School of Computer Science and Engineering, Nanjing University of Science and Technology;

【通讯作者】 王永利;

【机构】 南京铁道职业技术学院信息管理中心南京理工大学计算机科学与工程学院

【摘要】 为了精准地识别合成孔径雷达(Synthetic aperture radar, SAR)图像中的不同目标,该文提出了一种结合最近邻和拓展稀疏表示的分类(Nearest neighbor and extended sparse representation classification, NNSRC)方法。首先对图像进行预处理,抑制斑点噪声,归一化图像以保持有用的信息;接着采用二维主成分分析法来提取图像的特征向量,根据识别能力选取特征向量;最后判断SAR图像目标的类别。NNSRC方法拓展了稀疏表示模型,有效解决已有SAR图像识别方法处理噪声、遮挡问题效果差,以及求解稀疏系数速度慢的问题。标准SAR数据集上的实验结果显示,该文提出的NNSRC方法识别率达92.9%,与其他方法综合对比表明,NNSRC方法具有更加良好的性能。

【Abstract】 In order to accurately recognize different targets in synthetic aperture radar(SAR)images, a target classification method, called NNSRC,which combines nearest neighbor and extended sparse representation is proposed in this paper. Firstly, the image is preprocessed to suppress speckle noise and normalized to keep useful information. Secondly, the feature vector of the image is extracted by two-dimensional principal component analysis, and the feature vector is selected according to the recognition ability. Finally, the category of SAR image target is judged. NNSRC extends the sparse representation model, and effectively solves the problems of existing SAR image recognition methods, such as poor effect in processing noise and occlusion, and slow speed of solving sparse coefficients. Experimental results on standard SAR data sets show that the NNSRC method proposed in this paper has a recognition rate of 92.9%. Compared with other methods, the NNSRC method has better performance.

【基金】 国家自然科学基金(61941113);南京市科技计划项目(201805036);信息系统工程重点实验室开放基金资助课题(05202004)
  • 【文献出处】 南京理工大学学报 ,Journal of Nanjing University of Science and Technology , 编辑部邮箱 ,2021年05期
  • 【分类号】TN957.52
  • 【被引频次】5
  • 【下载频次】180
节点文献中: 

本文链接的文献网络图示:

本文的引文网络