吉林大学学报(理学版)

• 数学 • 上一篇    下一篇

面向全自动三维扫描系统的多视角三维数据自动配准技术

李杨1, 许志闻1, 宋展2, 庞旭芳2   

  1. 1. 吉林大学 计算机科学与技术学院, 长春 130012;2. 中国科学院深圳先进技术研究所 智能设计与机器视觉研究室, 广东 深圳 518055
  • 收稿日期:2013-05-01 出版日期:2014-03-26 发布日期:2014-03-20
  • 通讯作者: 许志闻 E-mail:xuzw@jlu.edu.cn

Automated Multiview 3D Data Registration forFullyAutomated 3D Scanning System

LI Yang1, XU Zhiwen1, SONG Zhan2, PANG Xufang2   

  1. 1. College of Computer Science and Technology, Jilin University, Changchun 130012, China;2. Intelligent Design and Pattern Recognition Center, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen 518055, Guangdong Province, China
  • Received:2013-05-01 Online:2014-03-26 Published:2014-03-20
  • Contact: XU Zhiwen E-mail:xuzw@jlu.edu.cn

摘要:

针对利用高精度转台进行配准模型的顶部和底部与模型进行配准时效果不佳的问题, 提出一种面向全自动三维扫描系统的多视角三维数据自动配准方案. 对待扫描模型, 利用单轴高精度转台获取由不同角度扫描的点云数据, 先使用一种快速简单的方法求解转台中轴, 利用中轴将多片点云拼合得到模型主体; 再通过点云包围盒变换, 完成点云的初始配准, 并通过改进的ICP算法完成点云的精确配准. 实验结果表明, 该方法具有较好的操作性和较高的精度.

关键词: 三维拼接, ICP算法, 最小包围盒

Abstract:

This paper reports an automated multiview threedimensional data registration scheme for fullyautomated threedimensional scanning system. We used singleaxis precise turntable to help to get multiview point cloud and to integrate them together. First, we used a quick and easy way to get the three dimensional information of the turntable’s central axis, and then integrated the multiview point cloud to get model body via the axis. After that we completed the initial registration of point cloud with the help of minimum bounding box. At last we completed the precise registration using improved ICP algorithm. In this paper, we provided several suggestions for the imprecise registration problem between the top and the bottom of the model. At the end of this paper, experiment results are presented, and experiments show that the scheme has good maneuverability and high accuracy.

Key words: 3D integration, ICP algorithm, minimum bounding box

中图分类号: 

  • TP391.7