CAJ | 학술논문

由于SAR图像与光学图像的显著差异 ,光学图像的目标识别算法并不能应用到SAR图像中 ,因此研究SAR图像的目标识别具有重要的意义.传统的基于模型的SAR图像目标识别算法中将所有的特征同等看待 ,然而不同特征对于目标分类的贡献度可能差别很大.给不同的特征赋予不同的权重 ,可能会改变目标在特征向量空间中的相对位置 ,从而给出更合理的识别结果.采用S A R图像的纹理特征作为分类特征 ,在支持向量机分类算法中加入使用Relief F算法计算得到的特征权重.试验结果表明这种加权后的目标识别算法具有更高的目标识别率.
유우SAR도상여광학도상적현저차이 ,광학도상적목표식별산법병불능응용도SAR도상중 ,인차연구SAR도상적목표식별구유중요적의의.전통적기우모형적SAR도상목표식별산법중장소유적특정동등간대 ,연이불동특정대우목표분류적공헌도가능차별흔대.급불동적특정부여불동적권중 ,가능회개변목표재특정향량공간중적상대위치 ,종이급출경합리적식별결과.채용S A R도상적문리특정작위분류특정 ,재지지향량궤분류산법중가입사용Relief F산법계산득도적특정권중.시험결과표명저충가권후적목표식별산법구유경고적목표식별솔.
T arget recognition algorithms of optics images can't be applied to SAR images because of the prominent differ-ence betw een SAR images and optics images ,w hich makes the research to SAR image target recognition more important . T raditional recognition algorithms based on models pay same attention to all features by giving the same w eight ,w hereas different features may have quite different influence to the recognition results .T he relative positions of target projected to feature space may change by giving different features with different w eights ,w hich helps to get more appropriate rec-ognition results .This article uses texture features as the recognition features ,then the SVM algorithms is adopted with the feature w eights calculated by ReliefF algorithm .T he test results show that this weighted recognition algorithm has higher recognition rate .