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

作　　者：张涛[1,2] 李晓宇[1,2] 魏敏 张昭[4] 高艳艳 ZHANG Tao;LI Xiaoyu;WEI Min;ZHANG Zhao;GAO Yanyan(Beijing Key Laboratory of High Dynamic Navigation Technology,Beijing Information Science&Technology University;Key Laboratory of Modern Measurement&Control Technology,Ministry of Education,Beijing Information Science&Technology University;School of Mechatronic Engineering and Automation,Foshan University;Hebei Jingjinji Institute of Remanufacturing Industry Technology)

机构地区：[1]高动态导航技术北京市重点实验室,北京信息科技大学 [2]现代测控技术教育部重点实验室,北京信息科技大学 [3]佛山科学技术学院机电工程与自动化学院 [4]河北京津冀再制造产业技术研究院

出　　处：《仪表技术与传感器》2024年第11期93-98,115,共7页Instrument Technique and Sensor

基　　金：国家自然科学基金重大科研仪器项目(52227804);国家自然科学基金面上项目(52274003);北京市教育委员会科学研究计划项目(KM202111232004)。

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。In response to the challenges of significant difficulty in three-dimensional reconstruction,low accuracy of dimension measurement,and heavy reliance on manual labor for waste parts during the remanufacturing process,a passive binocular vision-based method on three-dimensional reconstruction for waste parts was proposed.Initially,photographs of waste parts were captured using a binocular camera,and the Zhangs 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,incorporating 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.Experimental 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 requirements 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 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]