一种SAR图像特征提取和目标分类的新方法  

New Method for Synthetic Aperture Radar Images Feature Extraction and Target Classification

作　　者：李勇[1] 王德功[1] 常硕[1] 关春健[1] 

机构地区：[1]空军航空大学,吉林长春130022

出　　处：《现代防御技术》2013年第4期126-130,共5页Modern Defence Technology

摘　　要：提出了一种利用二维离散小波变换与核主成分分析结合对SAR图像进行特征提取的新方法。该方法对SAR图像进行三层小波分解后提取低频子带图像的核主成分分量作为目标的分类特征,利用支持向量机完成目标分类。采用MSTAR实测SAR目标数据集进行特征提取和分类实验,实验结果表明:该方法可以有效提高目标的正确识别率。A new method for synthetic aperture radar images feature extraction and target recognition based on Kernel PCA in wavelet domain and support vector machine is presented. After two-dimension wavelet decomposition of a SAR image, feature extraction is implemented by picking up Kernel principal component of the low-frequency sub-band image. Then, support vector machine is used to perform target recognition. Using MSTAR SAR data to experiment, results show that correctness of recognition is en-hanced obviously.

关 键 词：合成孔径雷达 二维离散小波变换 核主成分分析 支持向量机 自动目标识别 

分 类 号：TN958[电子电信—信号与信息处理]