Research On Laser Vision Sensor System For Arc Welding Robot

In the field of industrial arc welding,more and more factories have begun to rely on robots to implement semi-automatic and fully-automatic intelligent transformation of large-scale robots,which greatly improves welding production efficiency.Most of the traditional robotic are working by using the “teaching-online” mode,which has obvious effects in the early stage of factory intelligence.which has obvious effects in the early stage of factory intelligence.However,with the diversification and complication of processed parts,manual teaching has the problems of low efficiency and poor product consistency.In this paper,a laser vision platform is set up to achieve the trajectory planning of the welding seam through camera calibration,image preprocessing,and feature point extraction to control the robot to perform welding with high quality and efficiency.Firstly,building an arc welding robot welding system and a laser vision system.Construct the body,control cabinet and welding equipment of the welding robot welding system;select the hardware of the laser vision system,select the parameters of the camera,lens,laser generator,filter,and image acquisition card according to the technical requirements to obtain a clear laser stripe image.Secondly,the camera and hand-eye calibration theory are studied,and the internal parameter matrix and the hand-eye matrix are obtained by calibration.Based on Zhang's calibration theory,the calibration template was used to capture the calibration plates in 12 different locations in space and 12 pictures taken by different attitude cameras were obtained.The improved MATLAB toolbox is used to optimize the L-M gradient descent method to obtain the camera's internal parameter matrix,and then get the rotation of the hand eye matrix under each attitude.Finally,the 12-group matrix is averaged to obtain a hand-eye matrix.The average pixel calibration error of the camera is 0.8 pixels in the,which satisfies the accuracy requirement of the calibration accuracy within 2 pixels.Finally,this paper establishes the image preprocessing process,analyzes and compares the effects of different image processing welds,and obtains optimized image preprocessing algorithm parameters.The thinning algorithm and Hough transform are used to obtain the pixel coordinates of the weld center feature point.Programming to achieve grayscale,linear grayscale changes,grayscale nonlinear gamma conversion,image binarization,image smoothing and sharpening,extraction of the coordinates of the coordinates of the central pixel of the weld,through the matrix transformation to obtain space Three-dimensional coordinates.The paper verifies the stability of the laser vision sensor system through experiments,and compares the performance of laser vision welding and manual teaching.respectively.The X and Y direction accuracy increased by 23.9% and 30.3% respectively.efficiency is increased by 35.7%.This paper focuses on the design of the hardware of the laser vision system,vision system calibration,image preprocessing,weld feature extraction and three-dimensional coordinate transformation.Finally,a large number of experimental data verify the accuracy and stability of the system.The results show that the deviation of the system is within the 1.2mm error allowed by industry.This study has important practical value for improving the welding quality and efficiency of arc welding robots.