基于天绘一号卫星三线阵影像立体匹配的自动云检测  

Automatic Cloud Detection Based on Three- line Array Image Stereo Matching of TH- 1 Satellite

作　　者：胡振龙 吴腾[2] 张勇[2] 

机构地区：[1]中国天绘卫星中心,北京102102 [2]武汉大学遥感信息工程学院,湖北武汉430079

出　　处：《测绘科学与工程》2016年第3期12-18,共7页Geomatics Science and Engineering

摘　　要：有云场景对立体影像的自动地形提取带来很大影响，检测云区、利用无云区域的影像进行立体匹配获得DSM（Digital Surface Model），也是DSM自动化生产必须考虑的一个问题。本文基于云一般与地表分离的知识，利用密集匹配的DSM数据和参考DEM数据，结合高程与影像灰度信息提取出云区域，同时解决了云区域自动检测和地形提取。通过天绘一号卫星影像数据进行试验，无雪情况下云提取的正确率能达到95％；自动匹配的DSM也能有效地剔除云等因素的干扰，利用已有DEM数据进行修补，提高了成果质量。Cloud scene influences the automatic terrain extraction from stereo images greatly. Cloud detection and DSM ex- traction through stereo matching with cloudless images must be considered in DSM automated production. In this paper, based on the knowledge that cloud is far away from the earthg surface, the cloud area is extracted by using dense matching DSM and refer- ence DEM data and combining elevation data with image grey value information. In this way, the problems of cloud detection and automatic terrain extraction are solved simultaneously. TH - 1 satellite images are used to conduct test. It turns out that the accu- racy of cloud extraction can reach 95% when there is no snow, the automatic matching DSM can effectively prevent interference from cloud or other factors, and the quality of final result can be improved by modification with existing data.

关 键 词：立体影像 密集匹配 地形配准 云检测 

分 类 号：P231[天文地球—摄影测量与遥感]