Aiming at the problems of incomplete and incorrect water body extraction of large-scale high-resolution optical remote sensing image due to the complex background,the variation of water body and its surrounding environment,a semi-automatic extraction method for water bodies(lakes and waterways)is proposed.Firstly,the large-scale image is divided into several sub-image blocks with overlapping boundary by water body distribution,and the interested water bodies are marked in the form of rectangular boxes within them.Then,the spectral features of such water bodies are calculated using rectangular boxes as templates,and are further used as a matching term to achieve coarse extraction of sub image blocks of water bodies.Finally,the morphological features of the water body region are used to refine the rough extraction results.To validate the proposed method,GF-2 images containing different water bodies are tested using a local binary patterns and segmentation-based water body extraction method,and the results are compared and analyzed from qualitative and quantitative aspects.The results show the superiority of the proposed method.
关键词
水体/半/自动提取/大幅面/影像划分/高分辨率/形态特征
Key words
water body/semi-automatic extraction/large-scale/image division/high-resolution/morphological feature
维普
万方数据