基于多尺度联合分割的SAR图像电力线走廊变化检测  

SAR image power line corridor changing detection based on multi-scale joint segmentation

作　　者：吴晓东[1] 李瑛[1] 刘晓伟[2] 龚浩[2] 骆方[3] 

机构地区：[1]贵州电网公司输电运行检修分公司,贵州贵阳550002 [2]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 [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51-56,共6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：基于多尺度联合分割方法,提出一种SAR图像电力线走廊变化检测算法。该算法首先通过TurboPixels算法对多时相SAR图像电线塔走廊进行多尺度联合过分割,每个过分割区域即为一个"超像素";然后根据超像素区域的统计分布差异计算每个尺度上的差异度;最后融合多尺度差异,利用自动阈值提取最终的变化检测区域.实验结果表明该算法具有较为满意的变化检测效果.Based on multi-scale joint segmentation method, a SAR image changing detection algo- rithm over power corridor area is proposed in this paper. Firstly, through TurboPixels algorithm, multi-scale joint over-segmentation for multi-temporal SAR images over power corridor area is conducted. Each segment is considered as a ＂super-pixel＂. Then, the difference map based on the statistical distribution differences of super-pixel pairs is calculated. Finally, the final changed regions from the difference map are extracted by using an automatic threshold method. Experimental results show that the algorithm has a better changing detection performance.

关 键 词：合成孔径雷达 多尺度分割 变化检测 

分 类 号：TM835[电气工程—高电压与绝缘技术] TM764