摘要
从点云数据中提取建筑物轮廓是当前的一个研究热点,而现有算法大都需要先选取合适的种子点或不能很好地适应密度不均匀的点云数据。本文提出一种基于垂向密度快速提取点云数据建筑物矢量轮廓的方法,首先采用高程和面积阈值对滤波得到的非地面点分离出建筑物点云,然后基于垂向密度提取建筑物初始多段线,最后对初始多段线进行加权拟合提取建筑物规则化轮廓线。结果表明,基于垂向密度的点云建筑物轮廓提取方法无需其他辅助数据,且能较好地适应复杂地形,通过实验获取数据与实测数据对比分析可知,建筑物轮廓提取的准确度为90.98%、面积提取的准确度为94.32%、周长提取准确度为95.72%、位置精度均分误差为0.036 m,提取效果较好,可为点云数据的建筑物轮廓提取提供一种新方法。
Extracting building contours from point cloud data is a research hotspot at present,and most of the existing algorithms need to select suitable seed points or require point cloud data with even density.In this paper,we propose a method for quickly extracting building vector contours from point cloud data based on vertical density,firstly,the elevation and area thresholds are used to separate the non-ground points filtered to separate the building point clouds,then the initial polyline of the building is extracted based on the vertical density,and finally the regular contour of the building is extracted by weighted fitting of the initial polyline.The results show that the point cloud building outline extraction method based on vertical density does not need other auxiliary data and can adapt to complex terrain well,It can be seen through the comparison and analysis of the experimental data and the measured data.The accuracy of building outline extraction is 90.98%,the accuracy of area extraction is 94.32%,the accuracy of perimeter extraction is 95.72%,and the average position accuracy is 0.036 m,which provides a new method for the building outline extraction of point cloud data.
作者
蔡训峰
徐卓揆
袁齐
朱彬
Cai Xunfeng;Xu Zhuokui;Yuan Qi;Zhu Bin(School of Transportation Engineering,Changsha University of Science and Technology,Changsha 410114,China;Hunan Provincial Engineering Laboratory of Highway Geological Disaster Early Warning Spatial Information Technology,Changsha University of Science and Technology,Changsha 410114,China)
出处
《工程勘察》
2026年第2期70-75,共6页
Geotechnical Investigation & Surveying
基金
长沙理工大学公路地质灾变预警空间信息技术湖南省工程实验室开放基金资助项目(kfj180602).
关键词
LiDAR点云数据
矢量化
建筑物轮廓
垂向密度
多段线加权规则化
LiDAR point cloud data
vector quantification
building outlines
vertical density
polyline weighted regularization
