摘要
机载激光雷达能够快速获取地面三维结构信息,在城市三维建模、地形图测绘、电力线巡检等领域具有重要作用。点云滤波与分类作为点云数据处理中的重要环节,若提高其自动化程度及精度,则能够快速有效地从海量点云数据中提取出有用信息,如地面、植被、建筑物、管线等地物,因此本文基于城区机载点云数据,针对经典滤波方法进行对比分析。结果表明,相比于其他方法,布料模拟法能够有效地滤除非地面点,提供地形信息。随后在布料模拟法滤波的基础上,采用了随机森林对非地面点中常见的植被、建筑、管线等城市地物进行分类,有效地识别出各类地物,探索了适合城市的点云自动滤波及分类流程。
Airborne light detection and ranging(LiDAR)can rapidly acquire three-dimensional(3D)ground structure infor⁃mation,playing a crucial role in urban 3D modeling,topographic map surveying,power line inspection,and other fields.Point cloud filtering and classification,as important components of point cloud data processing,can significantly improve the automation and accuracy of extracting useful information,such as ground surfaces,vegetation,buildings,pipelines,and other objects from vast point cloud datasets.This paper analyzed and compared classical filtering methods using urban air⁃borne point cloud data.The results show that,compared to other methods,the cloth simulation method effectively filters out non-ground points and provides terrain information.Subsequently,based on the cloth simulation filtering,random forest is employed to classify common urban objects,such as vegetation,buildings,and pipelines,from non-ground points,effec⁃tively identifying various urban objects.The study explores an efficient automatic point cloud filtering and classification pro⁃cess suitable for urban environments.
作者
孙云飞
王华
孔俊元
SUN Yunfei;WANG Hua;KONG Junyan(The Second Surveying and Mapping Institute of Anhui Province,Hefei,Anhui 230601,China;AVIC Xingtu(Beijing)Information Technology Company Limited,Beijing 100080,China;Beijing Municipal Big Data Center for Landscape and Gardening,Beijing 100013,China)
出处
《北京测绘》
2025年第9期1351-1354,共4页
Beijing Surveying and Mapping
基金
安徽省自然资源厅2025年度自然资源科技项目(2025-K-18)
安徽省测绘局2024年度科技创新项目(2024-KJ-04)
自然资源部2024年度部省合作项目(2024ZRBSHZ152)
城市空间信息工程北京市重点实验室开放基金(20220102)。
关键词
布料模拟法
随机森林
激光雷达
点云滤波与分类
cloth simulation method
random forest
light detection and ranging(LiDAR)
point cloud filtering and classification
