吴天宝, 夏靖波, 黄玉燕.基于SVM和SRC级联决策融合的SAR图像目标识别方法[J].河南理工大学学报（自然科学版）,2020,39(4):118-124.

WU T B, XIA J B, HUANG Y Y.Target recognition method of SAR images based on cascade decision fusion ofSVM and SRC[J].Journal of Henan Polytechnic University(Natural Science) ,2020,39(4):118-124.

基于SVM和SRC级联决策融合的SAR图像目标识别方法

吴天宝1, 夏靖波1, 黄玉燕2

1.厦门大学嘉庚学院信息科学与技术学院，福建厦门 361005;2.集美大学轮机工程学院，福建厦门 361021

摘要:提出基于支持向量机（support vector machine ， SVM ）和稀疏表示分类（sparse representation-based classification ， SRC ）级联决策融合的合成孔径雷达（synthetic aperture radar，SAR ）图像目标识别方法。首先，采用SVM对测试样本进行分类，根据各个训练类别输出的后验概率，采用门限判决法选取其中具有高置信度的候选类别；其次，基于候选训练样本构造字典，对测试样本进行SRC分类；最后，采用线性加权融合SVM和SRC的决策值，获得更为可靠的识别结果。SVM的预筛选分类有效降低了 SRC中的字典规模，从而提高其分类效率，同时， SRC具有的噪声、遮挡稳健性也可以补充SVM在此方面的不足。因此，提出的方法可以有效综合 SVM和SRC的优势，提高最终的识别性能。采用MSTAR数据集进行识别实验，结果验证了本文方法的有效性。

关键词:合成孔径雷达;目标识别;级联决策融合;支持向量机;稀疏表示分类

doi:10.16186/j.cnki.1673-9787.2020.4.17

基金项目:福建省自然科学基金资助项目（2018J01101，600002/KL51828 ）；福建省中青年教师教育科研项目（JT180795 ）

收稿日期:2019/11/28

修回日期:2020/01/09

出版日期:2020/07/15

Target recognition method of SAR images based on cascade decision fusion ofSVM and SRC

WU Tianbao1, XIA Jingbo1, HUANG Yuyan2

1.School of Information Science and Technology， Xiamen University Tan Kah Kee College， Xiamen 361005，Fujian，China;2.Marine Engineering Institute，Jimei University，Xiamen  361021，Fujian，China

Abstract:A synthetic aperture radar（ SAR ）target recognition method was proposed based on cascade decision fusion of support vector machine （ SVM ）and sparse representation-based classification（ SRC ）. Firstly，SVM was employed to classify the test sample. Based on the posterior probabilities corresponding to different training classes，a threshold judgement algorithm was used to select those classes with high reliabilities. Secondly，the selected training classes were used to establish a dictionary to further classify the test sample by SRC. Finally，a linear fusion strategy was adopted to combine the results from SVM and SRC to obtain reliable recognition results. The prescreening by SVM could effectively decrease the scale of the dictionary for SRC thus improving the classification efficiency. Meanwhile，SRC had good robustness to conditions like noise corruption and occlusion ，which could complement the shortages of SVM. Therefore， the proposed method could effectively combine the advantages of SVM and SRC to enhance the final recognition performance. The MSTAR dataset was used to conduct target recognition experiments and the results showed the effectiveness of the proposed method.

Key words:synthetic aperture radar;target recognition;cascade decision fusion;support vector machine;sparse representation-based classification

基于SVM和SRC级联决策融合的SAR图像目标识别方法_吴天宝.pdf