测绘学报 ›› 2020, Vol. 49 ›› Issue (6): 757-766.doi: 10.11947/j.AGCS.2020.20190142

• 摄影测量学与遥感 • 上一篇    下一篇

多尺度区域生长点云滤波地表拟合法

詹总谦1, 胡孟琦1, 满益云2   

  1. 1. 武汉大学测绘学院, 湖北 武汉 430079;
    2. 钱学森空间技术实验室, 北京 100094
  • 收稿日期:2019-05-09 修回日期:2020-04-06 出版日期:2020-06-20 发布日期:2020-06-28
  • 通讯作者: 胡孟琦 E-mail:humengqi@whu.edu.cn
  • 作者简介:詹总谦(1978-),男,副教授,研究方向为数字摄影测量与计算机视觉。E-mail:zqzhan@sg.whu.edu.cn
  • 基金资助:
    国家自然科学基金(61871295)

Multi-scale region growing point cloud filtering method based on surface fitting

ZHAN Zongqian1, HU Mengqi1, MAN Yiyun2   

  1. 1. School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China;
    2. Qian Xuesen Laboratory of Space Technology, Beijing 100094, China
  • Received:2019-05-09 Revised:2020-04-06 Online:2020-06-20 Published:2020-06-28
  • Supported by:
    The National Natural Science Foundation of China (No. 61871295)

摘要: 针对经典的迭代三角网加密算法(PTD)过度侵蚀地形、误差累积的问题,提出了一种基于区域生长的多尺度滤波方法。该方法引入了金字塔策略建立不同层次的点云结构,以上层种子点为基准对下层种子点进行处理:先通过不规则三角网滤除非地面点,然后依据局部地形设置动态阈值,以表面拟合区域生长算法增长受侵蚀的地面种子点,循环迭代逐渐逼近真实地面。通过对ISPRS提供的15个基准数据集进行测试,第Ⅰ、Ⅱ类误差以及总误差分别为2.40%、3.67%、2.84%,Kappa系数为93.74%。结果表明,该算法具有更强的性能,可以获得理想的地面模型。

关键词: 点云滤波, 数字高程模型, 三角网, 数据金字塔, 区域生长

Abstract: Aiming at the problem of over-erosion and type II error accumulation by progressive triangle irregular network(TIN) densification(PTD), a multi-scale filtering method based on region growing is proposed. This method introduces pyramid strategy to establish different levels of point cloud, in which the low-level seed points are processed based on the high-level seed points. In the filtering process, the non-ground points are filtered by PTD first, and then the eroded ground seed points are compensated by the surface-fitting region growing algorithm with dynamic threshold determinated by local terrain, ultimately the real ground surface is gradually approached in loop iteration. By testing the 15 benchmark data sets provided by the ISPRS, Type I error, Type II error, Total error and Cohen’s kappa coefficient are 2.40%, 3.67%, 2.84% and 93.74% respectively, which shows that the proposed method has better performance to obtain the ideal ground model.

Key words: point cloud filtering, digital elevation model, triangle irregular network, data pyramid, region growing

中图分类号: