改进半全局匹配的高分卫星影像DSM提取方法  

A method for extracting DSM from high resolution satellite images on the basis of the improved semi-global matching

作　　者：邹思远 邹峥嵘[1,2] 潘红播[1,2] 李冰川 ZOU Siyuan;ZOU Zhengrong;PAN Hongbo;LI Bingchuan(School of Geosciences and Info-physics,Central South University,Changsha 410083,China;Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitoring,Ministry of Education,Central South University,Changsha 410083,China)

机构地区：[1]中南大学地球科学与信息物理学院,长沙410083 [2]中南大学有色金属成矿预测与地质环境监测教育部重点实验室,长沙410083

出　　处：《测绘科学》2020年第7期113-119,共7页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41701538);高分遥感测绘应用示范系统(二期)项目(42-Y30B04-9001-19/21)。

摘　　要：传统半全局匹配算法在提取高分卫星立体像对DSM时,匹配结果受正则化参数的影响较大,且在阴影、低纹理、重复纹理和视差跳变等难以匹配区域存在较多的视差空洞。针对上述问题,该文提出了一种改进半全局匹配的高分卫星影像DSM提取方法:在代价聚合过程中根据当前层金字塔影像的视差搜索范围和代价矩阵自动确定正则化参数以增强匹配结果的稳定性,同时构建金字塔聚合以提高匹配点的数量;在此基础上,利用基于快速双边算子的优化策略优化视差图以填充视差空洞。结果表明:本文方法生成的DSM在建筑物边缘更为精细,同时能够较好地恢复难以匹配区域的高程值,且具有更高的精度水平。Aiming at the problem that when the traditional semi-global matching was used to extract the DSM from high-resolution satellite stereo pairs,the result was greatly affected by the regularization parameters,and had invalid disparity in the areas of shadow,low texture,repeated texture and disparity transition.Therefore,this study proposed a DSM extraction method based on the improved semi-global matching.Firstly,in the cost aggregation process,the regularization parameters were automatically determined according to the disparity search range and the cost matrix under the current pyramid image to enhance the stability of the matching results,and the pyramid aggregation was constructed to increase the matching points.On this basis,the improved fast bilateral solver optimized the disparity map to fill the holes.The results showed that the DSM generated by the proposed method was more elaborate at the edge of the building,higher restoring ability of the elevation value of the invalid disparity area,and higher precision.

关 键 词：高分辨率遥感影像 半全局匹配 正则化参数 视差图 数字表面模型 

分 类 号：P237[天文地球—摄影测量与遥感]