摘要
该文基于高频区目标散射中心理论分析了高分辨率SAR图像机动目标和自然地物后向散射特性的差异,探讨了两类目标纹理现象产生的机理,并在此基础上,分别基于局部统计量和分形模型提取机动目标的纹理特征,给出了鉴别特征优选方法。文中利用MSTAR的车辆目标实测数据检验了该文计算的纹理特征,给出纹理特征优选结果以及各纹理特征鉴别的性能,结果表明该文提取的纹理特征具有较好的鉴别性能,能消除大部分自然地物产生的虚警。
Based on the theory of scattering center, the difference of back-scattering characteristics between the mobile targets and natural terrains in high-resolution SAR imagery is investigated, and the principles of texture features of the two kind targets are discussed in this paper. And then, the texture features are extracted by using respectively the local statistics and fractal models, and the method of selecting the best features is presented. The real vehicle target SAR image data in MSTAR database are used to test the texture features, the best features are selected and the discriminating performances of those features are shown. The results show that those features are good and can be used to eliminate the most false alarms of natural terrains.
出处
《电子与信息学报》
EI
CSCD
北大核心
2008年第12期2809-2812,共4页
Journal of Electronics & Information Technology
关键词
SAR图像
机动目标鉴别
纹理特征
感兴趣区域
SAR imagery
Mobile target discrimination
Texture feature
Region of interest