基于Hausdorff距离的城区高分辨率SAR图像配准方法研究  

Registration of High Resolution SAR Imagery in Urban Areas Based on Hausdorff Distance

作　　者：徐旭[1,2] 张风丽[1] 王国军[1,2] 邵芸[1] 

机构地区：[1]中国科学院遥感与数字地球研究所,北京100101 [2]中国科学院大学,北京100049

出　　处：《遥感信息》2014年第3期73-77,87,共6页Remote Sensing Information

基　　金：中国科学院知识创新工程重要方向项目(KZCX2-EW-320);国家自然科学基金项目(41001213);国家863计划(2011AA120403)

摘　　要：图像配准是图像融合、变化检测和目标跟踪等的基础。针对城区高分辨率SAR图像的特点和特征点选取的困难性,提出了一种基于Hausdorff距离的城区SAR图像自动配准方法。首先运用Otsu图像分割与Canny算子相结合的方法对图像中的道路或河流边缘特征点进行提取,再以Hasudorff距离为相似性测度对提取的特征点集进行匹配。基于TerraSAR-X图像的实验结果表明,运用该配准算法无需提取的特征点间一一对应,有效降低了特征提取的要求,并取得了较好的配准结果。Image registration is the base of image fusion,change detection and target tracking.Considering the characteristics of high resolution SAR image in urban areas and the difficulties of feature point extraction,an automatic registration method based on Hausdorff distance was proposed.First,edge points of roads or rivers were extracted using Otsu image segmentation and Canny operation.Then,the extracted feature points were matched with Hausdorff distance as the similarity measurement.Experiment based on TerraSAR-X images showed that good results can be achieved and there were less restrict requirements on the selected feature points.

关 键 词：图像配准 城区 高分辨率SAR HAUSDORFF距离 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]