基于单目立体视觉的焊缝粗定位方法  

Coarse Positioning Method of Weld Seam Based on Monocular Stereo Vision

作　　者：陆苗 王化明[1] 邰凤阳 朱雄伟 易文韬 LU Miao;WANG Huaming;TAI Fengyang;ZHU Xiongwei;YI Wentao(College of Mechanical and Electrical Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China)

机构地区：[1]南京航空航天大学机电学院,江苏南京210016

出　　处：《机械制造与自动化》2021年第6期207-210,共4页Machine Building & Automation

摘　　要：针对传统焊缝粗定位方法存在过程繁琐、适应性差等问题,提出一种基于单相机的立体视觉测量技术来实现焊缝粗定位。沿着直线或圆弧焊缝布置发光条削弱周围环境因素影响,控制单目相机以两个不同位姿采集图像,经灰度质心法提取发光条图像中心线,基于立体视觉模型重建焊缝特征点和直线或圆弧方程,获取焊缝在机器人基坐标系下的粗定位信息。实验表明:焊缝点坐标平均误差约为1.27 mm,能够在保证精确度的情况下实现对不同场景下的焊缝粗定位。As to the problems of traditional welding seam coarse positioning methods,such as cumbersome process,poor adaptability etc.,a method based on a monocular camera binocular vision measurement technology to realize weld coarse positioning is proposed.Light strips are arranged along the straight or arc welds to weaken the influence of surrounding environmental factors.The monocular camera is controlled to collect images with different poses.Upon the center line of the light strip images being extracted by the gray centroid algorithm,the feature points and straight or arc equations of the weld are reconstructed based on the stereo vision model to obtain the rough positioning information of the weld in the base coordinate of robot.The experiments show that the rough coordinate error of weld points is about 1.27 mm,which can realize the rough positioning of the weld under different scenes with guaranteed accuracy.

关 键 词：单目立体视觉 标定 空间点线重建 粗定位 

分 类 号：TG409[金属学及工艺—焊接]