Weld Seam Image Processing And Recognition Based On Machine Vision

At present,welding robots mainly implement re-welding of workpieces by dragging and teaching.This method is cumbersome and limited by the accuracy of the placement of the workpiece,which will result in uneven welding quality and difficult to complete welding tasks efficiently and accurately.In response to this problem,this paper builds and designs a laser vision platform to endow the welding robot with visual perception ability.It is of great significance and value to help welding industrial robots to realize welding automation by enabling them to accurately identify the spatial location of the weld.The type of weld identified in this topic is a V-shaped groove weld with a thickness of 10 mm and an angle of90°.Firstly,the design and construction of laser vision platform for weld image acquisition are completed.Secondly,the calibration experiment is carried out to solve the transformation relationship between each link.Finally,a set of weld image processing methods are designed to extract the characteristic points of the laser stripes of the V-shaped weld image.And they are converted into three-dimensional coordinates in the base coordinate system of the welding industrial robot.The main research contents of this paper are as follows:(1)Complete the selection of the main components of the laser vision platform and its design and construction.First of all,the main components of the laser vision platform were selected,including the Daheng Mercury MER-500-14 GC camera with Gig E interface,the M1214-MP2 lens produced by Computar,the FU brand wavelength 650 nm infrared laser generator and the 650 nm narrowband infrared filter.Then,combined with the actual needs of the welding robot,the reverse installation method was selected to fix the position relationship between the industrial camera and the linear laser generator.Solid Works is applied to design the installation jig for connecting industrial robot welding gun and the housing for fixing industrial camera and infrared laser generator.Finally,the design and construction of the laser vision platform are completed.(2)The laser vision platform designed and built in this project is calibrated,and the calibration parameters of each part of the laser vision platform are calculated.Firstly,using Matlab's camera calibration software to obtain the camera's internal parameter matrix and radial and tangential distortion parameters.Then,the Steger algorithm based on the Hessian matrix is introduced to achieve sub-pixel precision positioning of the center of the intersection of the line laser plane and the calibration plate.And the extracted 29,020 point pixel coordinates are transformed to the camera coordinate system.The least squares method is applied to compute the laser plane equation under camera coordinates.Finally,the “Eye-inHand” hand-eye system is calibrated by changing different robot poses,and the hand-eye conversion matrix from the camera to the end of the industrial robot is obtained.(3)Complete the processing and recognition of the weld image obtained by using the laser vision platform,and display the processed results on the designed graphical user interface in real time.Firstly,analyze the characteristic information of the V-shaped weld,which is identified.And clarify the target of weld image processing and identification.Then,complete the pre-processing of the welding seam image such as distortion,graying,filtering and denoising.Moreover,using the Steger algorithm to obtain the center line of the welding seam image,and filter it for denoising,linear interpolation,cubic B-spline smoothing,etc.Further,using the angle analysis method to detect the feature points of the weld image and convert them into world coordinates.Finally,Qt developed by Qiqu Company is used to complete the design of the graphical user interface,and the results of the weld image processing are displayed in real time.When the ROI of the region of interest of the weld image is set to be small,the weld image processing algorithm designed in this topic can make the processing time of each weld image about 80 ms,which basically meets the requirements for the processing speed of the weld image.(4)Designed the welding seam three-dimensional information acquisition experiment to verify the effectiveness of the laser vision platform designed in this project,the accuracy of the laser platform calibration and the correctness of the welding seam image processing and recognition algorithm.Firstly,the calibration of the welding gun coordinate system of the welding industrial robot is completed to prepare for obtaining the actual three-dimensional coordinates of the weld feature points.Then,the accuracy of the transformation of the feature points of the weld image into the three-dimensional world coordinates and the detection of the weld groove width was verified respectively.The experiment showed that the transformation error of the three-dimensional world coordinates of the weld image identification feature points was within 1.3mm,and the error of the detection of groove width is within 0.65 mm.It can be concluded that within a certain range,the longer the exposure time,the more accurate the recognition of image feature points.