基于分水岭变换的高分辨率遥感影像分割  

High-resolution remote sensing image segmentation based on watershed transform

作　　者：缪晓冬 陈曾敏 MIAO Xiaodong;CHEN Zengmin(Nantong Surveying and Mapping Institute Company Limited,Nantong Jiangsu 226000,China)

机构地区：[1]南通市测绘院有限公司,江苏南通226000

出　　处：《北京测绘》2022年第12期1690-1696,共7页Beijing Surveying and Mapping

摘　　要：针对原始分水岭分割在高分辨率遥感影像上严重的过分割现象,本文以基于边缘距离变换的形态学标记及光谱标记控制的分水岭算法为基础,结合形态学混合开闭滤波操作和基于巴氏系数和灰度共生矩阵的区域合并方法,提出了一种改进的分水岭分割算法。将本文算法与易康软件中的多尺度分割算法进行定量、定性的对比试验,分析表明本文算法在保留边缘信息的情况下,有更好的分割效果。In view of the serious over-segmentation phenomenon of the original watershed segmentation on high-resolution remote sensing images,this paper proposed an improved watershed segmentation algorithm based on the morphological marker and spectral marker-controlled watershed algorithm with the edge distance transformation,combined with the morphological hybrid open-close filtering operation and the region merging method with the Pasteur coefficient and gray level co-occurrence matrix.The quantitative and qualitative comparative tests between the proposed algorithm and the multi-scale segmentation algorithm in eCognition software showed that the proposed algorithm had better segmentation effect while preserving the edge information.

关 键 词：高分辨率遥感影像 分水岭分割 边缘检测 形态学标记 光谱标记 

分 类 号：P237[天文地球—摄影测量与遥感]