基于改进形态学标记分水岭算法的城镇违法建筑提取研究  

Research on the extraction of illegal buildings in cities and towns based on the improved morphological marker watershed algorithm

作　　者：林琪军 UN Qijun(Ningbo Tianyi Surveying and Mapping Design Research Co.Ltd,Ningbo 315000,China)

机构地区：[1]宁波市天一测绘设计研究有限公司,浙江宁波315000

出　　处：《测绘科学与工程》2020年第5期24-28,共5页Geomatics Science and Engineering

摘　　要：基于高分辨率遥感影像,在形态学标记分水岭算法基础上改进Sobel边缘算子对梯度图像的计算,从而完成城镇违法建筑物的提取,通过与其它传统算法的识别提取效果相比较,结果表明:改进形态学标记分水岭算法的平均识别率最高,达到91%,平均漏分率最小,仅为9%,平均虚假检测率较小,仅为4%,具有较好的建筑物提取识别效果,可为城镇规划部门的管理工作提供指导和帮助。Based on the high-resolution remote sensing image and the morphological marker watershed algorithm,the Sobel edge operator is improved to calculate the gradient image to extract illegal buildings in cities and towns.Compared with other traditional algorithms,the average recognition rate of the improved morphological marker watershed algorithm is up to 91%,the average leakage rate is only 9%,and the average false detection rate is only 4%.This method is efficient to extract and identify buildings,and can provide guidance and assistance for the management of town planning departments.

关 键 词：分水岭算法 遥感影像 图像增强 SOBEL算子 梯度图像 

分 类 号：P237[天文地球—摄影测量与遥感]