测绘通报 ›› 2017, Vol. 0 ›› Issue (2): 6-9,53.doi: 10.13474/j.cnki.11-2246.2017.0038

• 学术研究 • 上一篇    下一篇

点云数据的多几何面片特征自动识别

石宏斌1, 殷义程2, 袁曼飞3   

  1. 1. 许昌学院城乡规划与园林学院, 河南 许昌 461000;
    2. 云南省测绘工程院, 云南 昆明 650033;
    3. 陕西铁路工程职业技术学院, 陕西 渭南 714000
  • 收稿日期:2015-12-08 修回日期:2016-07-07 出版日期:2017-02-25 发布日期:2017-03-01
  • 作者简介:石宏斌(1981-),男,博士,讲师,主要从事激光点云数据分割、特征提取、自动配准等方面的研究。E-mail:hbshi@whu.edu.cn
  • 基金资助:

    河南省高等学校重点科研项目(17B420001);国家自然科学基金(41301429)

An Automatic Method for Point Cloud Multi-Geometry Feature Recognition

SHI Hongbin1, YIN Yicheng2, YUAN Manfei3   

  1. 1. School of Urban-rural Planning and landscape Architecture, Xuchang University, Xuchang 461000, China;
    2. Surveying and Mapping Engineering Institute of Yunnan Province, Kunming 650033, China;
    3. Shaanxi Railway Institute, Weinan 714000, China
  • Received:2015-12-08 Revised:2016-07-07 Online:2017-02-25 Published:2017-03-01

摘要:

针对人工构筑物丰富的面片特征,提出一种基于局部采样优化的多种几何面片(平面、柱面、球面)点云数据分割方法。该方法首先利用三维格网建立点云数据的空间划分,然后根据随机采样点确定局部格网单元,在格网单元内部拟合多种几何模型,通过局部打分确定局部候选模型(集),利用统计推断估算候选模型集的全局打分区间,最终获得当前最优模型及其一致集,从而实现点云数据分割。试验结果表明:该方法能够对富含规则几何特征的人工构筑物进行有效分割。

关键词: 地面激光三维扫描, 分割, 局部采样, 统计推断, 法向约束

Abstract:

A segmentation method is proposed in this paper,based on local sample for point cloud with rich geometry features(plane,cylinder,sphere). The proposed method firstly divides point cloud by 3D grid. Then a subcell is determined by a random sample point, in which multi-primitive models are fitted, and the candidate models are got by its local score. Then the global scores are determined by statistic inference. Hence, the best model and its consensus set can be selected, and a segmentation can be made on point cloud. The experimental results show that our method can efficiently segment man-made structures rich in regular geometry shape.

Key words: terrestrial laser 3D scanning, segmentation, local sample, statistics inference, normal vector constrain

中图分类号: