测绘学报 ›› 2023, Vol. 52 ›› Issue (10): 1724-1737.doi: 10.11947/j.AGCS.2023.20220371

• 摄影测量学与遥感 • 上一篇    下一篇

基于多特征聚类的复杂环境机载点云层次滤波方法

郭娇娇1,2, 陈传法1,2, 姚喜3, 刘妍1,2, 刘雅婷1,2, 刘盼盼1,2   

  1. 1. 山东科技大学测绘与空间信息学院, 山东 青岛 266590;
    2. 山东省基础地理信息与数字化技术重点实验室, 山东 青岛 266590;
    3. 山东省水利勘测设计院有限公司, 山东 济南, 250013
  • 收稿日期:2022-06-06 修回日期:2023-04-03 发布日期:2023-10-31
  • 通讯作者: 陈传法 E-mail:chencf@sdust.edu.cn
  • 作者简介:郭娇娇(1997-),女,硕士生,研究方向为点云数据处理。E-mail:Jiaojiao_guo@163.com
  • 基金资助:
    国家自然科学基金(42271438);山东省自然科学基金(ZR2020YQ26);山东省高等学校青创科技支持计划(2019KJH007)

A multi-feature clustering-based hierarchical filtering method for airborne LiDAR point clouds in complex landscapes

GUO Jiaojiao1,2, CHEN Chuanfa1,2, YAO Xi3, LIU Yan1,2, LIU Yating1,2, LIU Panpan1,2   

  1. 1. College of Geodesy and Geomatics, Shandong University of Science and Technology, Qingdao 266590, China;
    2. Key Laboratory of Geomatics and Digital Technology of Shandong Province, Qingdao 266590, China;
    3. Shangdong Survey and Design Institute of Water Conservancy CO., LTD, Jinan 250013, China
  • Received:2022-06-06 Revised:2023-04-03 Published:2023-10-31
  • Supported by:
    The National Natural Science Foundation of China (No. 42271438);Shandong Provincial Natural Science Foundation (No. ZR2020YQ26);Shandong Province Higher Educational Youth Innovation Science and Technology Program (No. 2019KJH007)

摘要: 机载激光雷达(LiDAR)点云滤波是点云数据处理的关键步骤,决定着后续派生品应用的精细程度。针对复杂场景区各种地物的交错性和多态性、地形的突变(断裂)性、相邻地物和地面点高程的相似性等导致的难以区分地物点和地面点瓶颈,本文提出了一种基于多特征聚类的点云层次滤波方法。本文方法首先耦合点云几何和物理信息进行多特征点云聚类,然后发展一种顾及地形断裂的地面点簇识别方法捕捉各类地面点,最后利用捕捉到的地面点构建初始地面参考面,并借助多尺度层次点云滤波方法进一步查找原始点云中的地面点。以4组地形复杂且建筑物和植被混杂区点云数据为试验数据,将本文方法与6种代表性滤波算法对比表明,本文方法的平均总误差最小、滤波性能最优、稳定性最高。

关键词: 点云滤波, 机载LiDAR, 点云聚类, 多特征

Abstract: Airborne LiDAR point cloud filtering is the key step in point cloud processing, and its computational accuracy significantly affects the granularities of subsequent applications. However, it is difficult for the existing filtering algorithms to effectively distinguish object points from ground points in complex areas. Thus, a multi-feature clustering-based hierarchical filtering method is proposed in this paper. The proposed method first performs multi-feature point cloud clustering based on the geometric and physical information of point clouds; then, a ground-cluster identification method was used to accurately capture ground points in the area with breaklines; finally, the ground reference surface was constructed through the initial ground points, and the multi-scale hierarchical filtering was employed to further identify missed ground points. The new method was used to handle the point clouds in four different areas, and the filtering results were comprehensively compared with six state-of-the-art filtering algorithms. Results show that the new method has the lowest average total error, the best filtering performance and the highest stability.

Key words: point cloud filtering, airborne LiDAR, point cloud clustering, multi-feature

中图分类号: