Research On Three-dimensional Information Of Weld Image Based On Machine Vision

The automation and intelligence of welding technology is the key guarantee for improving welding quality and reducing production costs.The rapid development of machine vision technology depends on the improvement of image sensor devices and intelligent image processing methods,now many welding robots and certain welding devices have a certain degree of Visual function.The vision-based measurement is a non-contact measurement.which carries a wealth of information,has high accuracy and strong anti-interference ability,making it possible to apply in the weld seam recognition measurement and weld path planning.In view of the shortcomings of the current weld detection,such as low accuracy of detection and low intelligence,this paper has conducted in-depth study of visual weld seam measurement technology and image processing technology of the weld seam,which uses the method of binocular vision measurement based on structured light before welding-Detection,which uses matlab-program to identify and handle welds.This paper mainly completes the following aspects:(1)Based on the structure and working principle of the line-structured light-based binocular vision measurement system,the hardware platform of the measurement system is established,and the selection of the hardware device is determined.The key is to determine the installation design of the camera and wire structure light,test and adjust the spatial position between the two,so that the overlay image of the weld image captured by the left and right cameras is better.(2)For binocular vision,comparing different camera calibration methods,a calibration method based on Prof.Zhang Zhengyou was used to calibrate the grid.Realizing the acquisition of the parameters inside and outside the camera,as well as the acquisition of distortion parameters.Realizing the transformation of feature information from two-dimensional space coordinates to three-dimensional space coordinates.(3)Through the pretreatment of the collected weld seam images,weld images with less interference are obtained.Obtained the centerline of the optical band of the weld,using feature points and straight line detection methods to extract the information of weld size and position.(4)Binocular matching is performed on the image information extracted by the left and right cameras,and feature point matching based on limit constraints is used.According to the results of the dual target determination,the three-dimensional coordinates of the weld feature points are calculated.Through the binocular vision measurement experiment,the measured width and depth of the weld seam are the same as the weld seam size information of the workpiece used,and the welding seam accuracy measurement requirements in the welding project are achieved.Through visually identifying and measuring the size and position information of the weld,and fitting the discrete space point to the space straight line,the fitting of weld seam is realized to facilitate automatic welding of special welding device.