• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2020, Vol. 56 ›› Issue (9): 215-222.doi: 10.3901/JME.2020.09.215

• 数字化设计与制造 • 上一篇    下一篇

扫码分享

序列图像约束的点云初始配准方法

孙殿柱1, 沈江华1, 李延瑞2, 林伟1   

  1. 1. 山东理工大学机械工程学院 淄博 255049;
    2. 西安交通大学机械工程学院 西安 710049
  • 收稿日期:2019-06-29 修回日期:2019-12-07 出版日期:2020-05-05 发布日期:2020-05-29
  • 通讯作者: 沈江华(通信作者),男,1995年出生,硕士研究生。主要研究方向为逆向工程、数字化设计与制造。E-mail:jianghuash@126.com
  • 作者简介:孙殿柱,男,1956年出生,博士,教授,博士研究生导师。主要研究方向为逆向工程、数字化设计与制造。E-mail:dianzhus@sdut.edu.cn;李延瑞,男,1979年出生,博士研究生。主要研究方向为逆向工程、数字化设计与制造。E-mail:liyanrui.m2@gmail.com;林伟,男,1994年出生,硕士研究生。主要研究方向为逆向工程、数字化设计与制造。E-mail:linwei@126.com
  • 基金资助:
    国家自然科学基金资助项目(51575326)。

Initial Registration Method of Point Cloud Based on Sequence Images

SUN Dianzhu1, SHEN Jianghua1, LI Yanrui2, LIN Wei1   

  1. 1. School of Mechanical Engineering, Shandong University of Technology, Zibo 255049;
    2. School of Mechanical Engineering, Xi'an Jiaotong University, Xi'an 710049
  • Received:2019-06-29 Revised:2019-12-07 Online:2020-05-05 Published:2020-05-29

摘要: 针对现有算法难以精确配准重叠区域较小的点云,提出一种结合二维序列图像的点云初始配准方法。基于移动式三维测量方法获取实物表面灰度图像及三维数据,根据透视投影原理对相机在点云局部坐标系中的位置进行定位,获取点云到对应相机坐标系的变换矩阵,并以灰度图像的特征点及其匹配点作为匹配对,通过重建序列图像对相机外参数予以全局优化,根据初始配准公式实现点云初始配准。试验结果表明,该算法可正确配准重叠区域较小的点云,且能显著提高配准过程的稳健性及配准效率。

关键词: 点云初始配准, 移动式三维测量, 运动法重建, 序列图像

Abstract: An initial registration algorithm for point clouds combining two-dimensional sequence images is proposed to solve the problem that the existing algorithms are difficult to accurately register the point clouds with small overlapping area. The gray-scale images and point cloud data of physical surface are obtained based on the mobile three-dimensional measurement method. It depends on the principle of perspective projection that the position of a camera in the local coordinate system of a point cloud is located, and a transformation matrix that transforms the point cloud to the corresponding camera coordinate system is obtained. The image feature points and their matching points are used as matching pairs, the external parameters of the camera are globally optimized by reconstructing the sequence images, and then the point cloud registration is realized according to the initial registration formula. The results of experiment show that the method can correctly register the point cloud with small overlapping area and significantly improve the robustness and efficiency of the registration process.

Key words: initial registration of point clouds, mobile three-dimensional measurement, structure from motion, sequence images

中图分类号: