基于双目视觉的废旧零件三维重建方法研究
Research on Three-dimensional Reconstruction Method for Waste Parts Based on Binocular Vision
张涛 1李晓宇 1魏敏 2张昭 3高艳艳3
作者信息
- 1. 高动态导航技术北京市重点实验室,北京信息科技大学;现代测控技术教育部重点实验室,北京信息科技大学
- 2. 佛山科学技术学院机电工程与自动化学院
- 3. 河北京津冀再制造产业技术研究院
- 折叠
摘要
针对再制造过程中废旧零件三维重建难度大和尺寸检测精度差、依赖人工等问题,提出了一种基于被动式双目视觉的废旧零件三维重建方法.首先,通过双目相机采集废旧零件的照片,采用张正友标定算法获取相机内外参数,并对左右图像进行极线校正;其次,利用改进的AD-Census算法获取视差图,包括基于正态分布的Census变换改进和基于最短臂长的自适应权重设置,以提高匹配精度;最后,通过三角测量原理将得到的视差图转化为三维点云图,实现对零件的三维重建,并进一步进行尺寸测量.实验结果表明,所提出的方法在废旧零件三维重建中取得了良好的效果,且在尺寸测量方面的绝对测量误差控制在0.1 mm以内,能够满足废旧零件尺寸测量的检测精度要求,适用于实际生产中废旧零件表面信息分析.
Abstract
In response to the challenges of significant difficulty in three-dimensional reconstruction,low accuracy of dimen-sion measurement,and heavy reliance on manual labor for waste parts during the remanufacturing process,a passive binocular vi-sion-based method on three-dimensional reconstruction for waste parts was proposed.Initially,photographs of waste parts were captured using a binocular camera,and the Zhang's method was employed to calibrate the camera parameters and rectify the left and right images along epipolar lines.Subsequently,an improved AD-Census algorithm was used to obtain the disparity map,in-corporating enhancements such as Census transformation based on normal distribution and adaptive weighting based on the shortest arm length to improve matching accuracy.Finally,the obtained disparity map was converted into a three-dimensional point cloud using the principle of triangulation,facilitating three-dimensional reconstruction of the parts and further dimension measurement.Experi-mental results demonstrate that the proposed method achieves good results in the three-dimensional reconstruction of waste parts,with an absolute measurement error controlled within 0.1 mm in dimension measurement,meeting the detection accuracy require-ments of waste parts dimension measurement.This method is suitable for surface information analysis of waste parts in practical production.
关键词
双目视觉/双目标定/立体匹配/三维重建/尺寸测量Key words
binocular vision/binocular calibration/stereo matching/three-dimensional reconstruction/dimensional measurement引用本文复制引用
出版年
2024
NETL
NSTL
万方数据