| With the rapid development of China’s industrial manufacturing,there is a higher accuracy requirement for product measurement,especially in the welding field of China,intelligent welding robots have improved welding accuracy and productivity to a certain extent due to their visual functions Therefore,research on the welding recognition technology of welding robot based on vision is of great significance to the development of the welding field in China.In this paper,aiming at the shortcomings of large truck frame welding,such as low measurement accuracy and low welding intelligence,binocular vision is applied to the welding robot.Before the welding robot welds the welding,the frame welding recognition technology is studied in depth.In Microsoft Visual Studio software,using C # language programming to realize the welding robot frame welding recognition system.The specific research contents are as follows:(1)The paper introduces the hardware environment for system construction and the prerequisite for system use is that the welding robot has reached the frame weld near the frame through the overall recognition of the macro environment.Based on this,this article uses structured light binocular vision technology to identify the frame weld.The main task of binocular vision is to simultaneously shoot the weld of the frame from different angles and measure the depth of the weld.According to the principle of binocular vision technology,using the Matlab toolbox to complete the calibration of the binocular camera,according to the camera parameters obtained through the principle of Bouguet operator epipolar geometry to complete binocular image correction.Because the binocular vision technology based on structured light can improve the accuracy of weld identification and measurement through a wealth of feature information,after comprehensively analyzing the characteristics of structured light,a coded structured light with repeated periodic combinations is designed.The measurement range of the measured object can also improve the accuracy of stripe feature extraction and decoding.(2)The thesis researches the weld recognition algorithm of the frame weld image.The images collected by the welding seam include images of strong light projection and structured light projection.For binocular images projected by strong light,after analyzing various edge extraction operators,an improved Canny operator is proposed for the edge of the weld contour.The experimental results show that the algorithm Edge detection is superior to other edge operators.In order to determine the pixel coordinates of the frame weld in the collected image,a method based on pixel neighborhood comparison is used to track the edge of the weld contour.In the stage of binocular visual stereo matching,an algorithm combining structured light decoding and extraction of the contour boundary of the binocular frame welding is proposed to complete the stereo matching,and the space coordinate of the welding contour is calculated according to the matching point.Finally,according to the paper proposed to adopt the method of global least squares singular value decomposition to fit the spatial data of the weld to identify the spatial position of the weld,and determine the start and end points of the weld.The results are compared with the actual measured values.The error is about ± 0.2mm.(3)The paper in Microsoft Visual Studio software,through C # language programming to achieve welding robot frame welding recognition system,the system mainly includes binocular collected frame welding image correction,welding contour boundary extraction,three-dimensional matching,spatial point calculation,spatial point data fitting and other algorithms.In addition,the interface design of the system is performed and the recognition effect is demonstrated,which proves the accuracy and feasibility of the welding seam recognition system of the welding robot frame studied. |
