Research On Noise Image Processing And Rapid Location And Seam Tracking Of Robot Welding Zone Based On Laser Vision

Robot intelligent welding that is an indispensable part of modern intelligent manufacturing,has the advantages of high efficiency,stable quality and improving working conditions,and it is widely used in automobile manufacturing,ship heavy industry,equipment manufacturing and other fields.At present,most of the robots used in actual welding production at home and abroad are"teaching-reappearing"robots.This type of robot has no adaptability to the deviation of teaching trajectory,does not have the function of initial welding position guidance and real-time tracking control of welding seam.At the same time,the actual welding conditions are complex and changeable,which increases the difficulty of robot initial position guidance and real-time seam tracking control.In this paper,based on laser vision sensing technology,noise image processing and fast positioning and seam tracking technology of robot welding area are studied.Sensors play an important role in the initial position guidance and real-time seam tracking control of robots.In this paper,a visual sensor hardware system is developed based on 3D printing technology.With the idea of modularization and integration,this paper develops a software system for robot initial position guidance and real-time seam tracking control.The experimental results show that the software and hardware system of the sensor meets the requirements of the real-time tracking control of the initial welding position and the welding seam in the actual welding process.Based on pinhole and three-dimensional stereo model,this paper calibrates the vision system,realizes the transformation from image pixel coordinates to space world coordinates,gives the detailed visual calibration parameters and their solving process,and analyses the calibration error and gives the error formula.The calibration accuracy of the vision system is0.6mm,which meets the requirements of actual welding location and seam tracking.According to the actual requirements of welding,a three-line search method is proposed for the type of straight-line weld,and the parameters are solved to carry out rapid guidance and positioning.Aiming at the common typical V-type,T-type and butt welds,this paper develops a software module to locate the initial welding points.Experiments show that the software module can work quickly and effectively.Because of the complexity of welding process,the seam tracking system based on laser vision sensing technology is often unstable due to noise interference.Aiming at several kinds of common welding images with strong noise,such as atypical welding seam,strong arc light and large spatter,this paper uses the method of fast image segmentation,convolution neural network feature area recognition and feature search technology to realize the stable recognition of welding seam features and increase the robustness of real-time seam tracking system.Based on the strong noise welding image processing method,the selection threshold of gray image in the process of image processing is increased from 0.5×10~7 to 1.5×10~7.The length of the welding image time is extended to 2 times as much as that of the original image,which significantly increases the stability and practicability of the real-time seam tracking system.Using self-designed visual sensing system,the validation experiments of T-type weld,V-type weld and butt weld were carried out.The initial positioning accuracy was 0.80 mm,0.93 mm and 0.9 mm,respectively.At the same time,the real-time tracking control experiments of T-type weld and V-type weld were carried out,and the tracking accuracy was 0.50mm and 0.62mm,respectively.The whole system meets the needs of actual robot welding production.