SAR图像球流形局部嵌入建模及其分类方法  

SAR image spherical local embedding modeling and classification method

作　　者：周乐意[1] 余文涛[1] 陈嘉宇[1] 孙洪[1] 

机构地区：[1]武汉大学电子信息学院,湖北武汉430072

出　　处：《信号处理》2013年第9期1163-1168,共6页Journal of Signal Processing

摘　　要：地物目标建模是合成孔径雷达(Synthetic Aperture Rader,SAR)图像解译和应用的关键技术之一。近年来,基于流形学习的建模方法得到发展,可望适用于依据微波散射机理成像的SAR图像建模。本文采用球流形嵌入(SLE)方法来实现SAR地物目标建模。该方法实质上是对SAR图像的高维描述或表达进行非线性降维,得到相应的低维流形结构,其分量就是SAR图像的本质特征,由于削弱了原始高维表达中的冗余信息,可用来更加精确地描述和解译地物目标,同时由于维数的降低,大大降低了计算复杂度。为验证其有效性,本文将其应用于SAR图像场景分类,采用简单的K最近邻(K nearest neighbor,KNN)分类器和支持向量机(Support Vector Machine,SVM)分类器。实验结果证明基于本文方法对SAR图像地物目标建模是有效的,有着良好的应用前景。Modeling for SAR target is one of key technologies for SAR image interpretation and application. Recently, modeling based on manifold learning develops so that it is applicable to model for SAR image which is based on the imaging mechanism of microwave scattering. In this paper, SLE method was used to model for SAR image. In essential, this meth- od did nonlinear dimensionality reduction on high dimensional SAR image data, capturing the corresponding low-dimension- al manifold structure indicating intrinsic features. It could be used to represent and interpret objects more precisely, due to weakening the redundant information of original high-dimensional data. Additionally, the reduction of dimensionality greatly reduced the computational complexity. To verify its effectiveness, this paper applied it on SAR image scene classification, using KNN classifier and SVM classifier. The experiment results demonstrated that our method was effective and had a good prospect.

关 键 词：合成孔径雷达图像 流形学习 球流形局部嵌入 建模及分类 

分 类 号：TN911.73[电子电信—通信与信息系统]