多尺度自适应加权与稀疏表示分类相结合的遥感目标识别  

Remote Sensing Object Recognition Based on Gabor Multi-scale Adaptive Weighting and Sparse Representation Classification

作　　者：李骥[1] 王艳然 王威[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院综合交通运输大数据智能处理湖南省重点实验室,长沙410114

出　　处：《小型微型计算机系统》2017年第9期2157-2160,共4页Journal of Chinese Computer Systems

基　　金：国防"九七三"重点基础研究项目(613XXX0301)资助

摘　　要：针对遥感图像中不同层次的空间结构差异及目标含有不同角度的旋转的情况,提出一种基于Gabor多尺度自适应加权与稀疏表示的遥感目标识别方法.首先对训练样本和待测样本进行Gabor小波变换,对各个方向的Gabor特征进行综合,使它们近似各向同性,根据各尺度特征包含信息量进行自适应加权求和并经过PCA降维求得融合特征,将原始的训练字典改为融合特征字典,从而使字典更加具有判别能力,提高识别率.实验表明,该方法对遥感图像目标识别具有较好的鲁棒性.Aiming at different levels of spatial structural differences and target with different Angle of rotation in remote sensing images,a remote sensing target recognition approach based on Gabor multiscale adaptive weighting and sparse representation classification is proposed.First,Gabor wavelet transform is performed on the training samples and testing samples.The Gabor features in each direction are synthesized to be approximate isotropic.Then,fusion features are achieved by adaptive weighting summation and PCA dimensional reduction according to the characteristics contained in each scale.Thus a fusion feature dictionary is obtained replacing the original training dictionary so that the dictionary can be more discriminative,and the recognition rate can be higher.Experiments show that the method is robust to the recognition of the remote sensing target.

关 键 词：Gabor多尺度 自适应加权 稀疏表示 融合特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]