基于多特征融合的SAR图像输电走廊变化检测  

Multiple features fusion for SAR image power line corridor changing detection

作　　者：刘晓伟[1] 孟岩[2] 龚浩[1] 周赞东[1] 骆方[3] 

机构地区：[1]国家电网公司电力科学研究院武汉南瑞有限责任公司,湖北武汉430074 [2]新疆电力公司,新疆乌鲁木齐830002 [3]武汉大学电子信息学院,湖北武汉430072

出　　处：《电力科学与技术学报》2012年第4期57-63,共7页Journal of Electric Power Science And Technology

基　　金：国家重点基础研究发展计划("973计划")(2009CB724507-3)

摘　　要：传统SAR变化检测使用单一描述来提取多时相SAR图像间的变化信息,没有充分挖掘图像中的多特征信息,导致复杂变化场景下算法检测精度不高.针对这一问题,提出一种基于多种特征融合的SAR图像变化检测方法.该方法首先对多时相SAR图像输电走廊区域进行多种特征提取,并选取合适SAR变化检测的特征,然后在多时相SAR图像中计算每种特征对应的差异图,最后从图像融合的角度分别使用主成分分析法(PCA)和证据推理理论(DST)对这些多特征对应的差异图进行融合并提取最终的变化检测结果.实验结果表明该方法具有较为满意的变化检测效果.The single descriptor is applied to extract change information between muti-temporal SAR images in traditional SAR change detection methods. It does not make full use of the image information, resulting in a drop in detection accuracy especially in complex scene. To solve this problem, a SAR image change detection method based on multiple features fusion was presented in this paper. Firstly, multiple features were extracted from multi-temporal SAR image power line corridor area, and the appropriate SAR change detection characteristics were selected. Then, the corresponding difference maps to each feature channel were further extracted. Finally, from the perspective of image fusion, the main component analysis （PCA） and evidence reasoning the- ory （DST） were separately used to fuse the difference maps corresponding to these features, and meanwhile the final changing detection results are extracted. Experimental results showed the proposed method has satisfactory effectiveness for changing detection.

关 键 词：合成孔径雷达 特征融合 变化检测 输电线路 

分 类 号：TM835[电气工程—高电压与绝缘技术] TM764