计算机与现代化

• 图像处理 • 上一篇    下一篇

基于单目视觉几何构建的三维扫描系统

  

  1. 1.西北农林科技大学信息工程学院,陕西杨凌712100;2.山西省运城市农业综合开发办公室,山西运城044000
  • 收稿日期:2014-03-20 出版日期:2014-06-13 发布日期:2014-06-25
  • 作者简介: 裴骄阳(1975-),男,山西运城人,西北农林科技大学信息工程学院硕士研究生,山西省运城市农业综合开发办公室经济师,研究方向:计算机应用技术和农业信息化; 通信作者: 张志毅(1974-),男,副教授,博士生导师,研究方向:计算机图形学,机器视觉,精密仪器设计; 袁聪聪(1988-),男,河南商丘人,硕士研究生,研究方向:计算机图形学,数字图像处理,光电仪 器测量系统设计。
  • 基金资助:
     国家高技术研究发展计划(863计划)资助项目(2013AA10230402); 中央高校西北农林科技大学基本科研项目(QN2013054)

A 3D Scanning System Based on Monocular Vision Geometry Construction

  1. 1.College of Information Engineering, Northwest A&F University, Yangling 712100, China;

     2.Yuncheng Comprehensive Agricultural Development Office of Shanxi, Yuncheng 044000, China
  • Received:2014-03-20 Online:2014-06-13 Published:2014-06-25

摘要:

为构建基于单目视觉的快速便捷式三维扫描系统,提出一种高精度的单目视觉几何投影的场景建模方法,并开发一种低成本高精度的三维扫描系统。首先,获取平面标定点的图像坐标,经
投影变换将其转换为摄像坐标系下的三维空间位置坐标,分别建立平移台面和底座的三维空间平面方程;其次,通过移动平移台面求取同名标定点的空间坐标,求解平移台面的平移向量,并通过落在平
移台面和底座上的激光线条求解激光平面;最后,提取图像中的激光光条中心点并将其变换为物体表面的三维点云数据。实验结果表明,投影变换求得的平面方程误差小于0.2%,扫描结果误差低于
0.05 mm。

关键词:  , 激光光学, 三维扫描, 单目视觉, 几何构建

Abstract:

In order to construct a convenient and fast 3D scanner based on monocular vision, a high-precision monocular visual geometric projection  of scene modeling
method is proposed and a high-precision 3D scanning system is developed. Firstly, we obtain the coordinates of calibration points by image processing, and convert them into 3D
camera coordinate system by the projective transformation, and then construct the plane equation of Mesa and Base. Secondly, we calculate translation vectors through moving Mesa
to obtain the translation vector, and get the laser plain equation by the camera center and the laser line which fell on the Mesa and Base. Finally, we extract the line center
of laser light strip, and obtain the 3D point cloud data of object surface after projection transformation and translation transform. Experiment results indicate that through
the geometry projection transformation method to construct the plane equation error is less than 0.2%, and the error of 3D scan results is less than 0.05 mm.

Key words: laser optics, 3D scanning, monocular vision, geometry construction