Research On Automatic Welding Method Based On Monocular Visual Image Crack Detection And Tracking

With the development of welding technology and the improvement of automated testing methods,the working intensity and working environment of welding workers have been effectively improved.Welding was realized by welding robot instead of manual work;the automated detection method was changed from traditional manual detection to mechanical vision detection;the traditional manual production was slowly transformed into advanced semi-automatic,fully automated production.Such changes not only had important theoretical significance,but also had a far-reaching impact on the development of automated welding based on visual and image processing.In this paper,a method of automatic welding for crack detection and tracking based on monocular vision was proposed.Combining with the combination of machine vision and image processing,the welding model based on industrial camera was established;in this paper,an image processing algorithm for weld characteristics was proposed and visual seam tracking system was designed.Mainly divided into the following aspects:(1)In the process of welding,the acquisition of the welding characteristics of the camera is interfered by the laser and other strong light.The mathematical model of image acquisition under strong light interference is put forward,and the original image feature under strong light interference is analyzed theoretically.(2)This paper analyzes the characteristics of welds collected,studies image enhancement algorithms,and proposes an image enhancement technique based on the combination of the dark primary and de-fog techniques,which is more superior to the traditional gray value equalization image enhancement method.(3)This paper analyses the traditional pixel level edge detection,proposes an edge detection method based on sub-pixel accuracy.Sub-pixel edge detection has the advantages of high positioning accuracy and good detection effect.(4)The paper is based on the calibration of HALCON visual software,using HALCON calibration assistant to achieve camera calibration and experimental analysis of image deformity.(5)The communication model of laser range finder and robot is established,using laser range finder to guide the welding torch and accurate positioning of the weld position.The experimental results verify the accuracy of the robot and laser ranging,and meet the requirements of high precision welding equipment.(6)Basing on the simulation of welding seam and the Visual Studio programming software,combining HALCON programming software system for weld detection and tracking,and robot communication,the paper puts forward the feasibility of detecting and tracking in monocular vision image based on crack automatic welding.