摘要
针对经典的建筑物边界正则化方法(如最小外接矩形法和角点拟合法)无法适用于高分辨率遥感影像中形状复杂的建筑区域,该文提出了一种基于感知编组的建筑物轮廓正则化处理方法。基于道格拉斯-普克算法对粗提取的建筑物进行多边形简化,利用关键点信息进行分组最小二乘直线拟合,在建筑物主导方向约束下利用感知编组算法进行各个线段的连接与重构。实验结果表明,该文正则化处理方法在建筑物轮廓的准确性、规整程度及最终精度均得到提高,能真实、准确地反映建筑物的真实形状。
Building boundary regularization is an important step in building extraction.Classical boundary regularization methods(such as minimum circumscribed rectangle method and corner fitting method)cannot be applied to building areas with complex shapes in high-resolution remote sensing images.This paper proposes a building contour regularization method based on building-aware grouping.The polygons of the roughly extracted buildings are simplified based on the Douglas-Puck algorithm;The hierarchical least squares line fitting is performed using the key point information;The perceptual grouping algorithm is used to connect each line segment under the constraint of the dominant direction of the building with refactoring.The experimental results show that the regularization processing method in this paper has improved the accuracy,regularity and final precision of the building outline,and can truly and accurately reflect the real shape of the building.
作者
解斐斐
顾宇超
霍志玲
XIE Feifei;GU Yuchao;HUO Zhiling(College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao,Shandong 266590,China)
出处
《测绘科学》
CSCD
北大核心
2023年第1期49-59,共11页
Science of Surveying and Mapping
基金
山东省自然科学基金项目(ZR2021MD026)
山东省高等学校科技计划项目(J18KA214)
关键词
正则化
高分辨率遥感影像
直线拟合
主导方向
感知编组
regularization
high-resolution remote sensing imagery
line fitting
dominant direction
perceptual grouping
作者简介
解斐斐(1983—),女,山东青岛人,讲师,博士研究生,主要研究方向为遥感图像处理、低空摄影测量。E-mail:xff@sdust.edu.cn
