Research And Implementation Of Welding Seam Positioning And Tracking System Based On Machine Vision

Welding,as a significant processing technology,has been widely used in various manufacturing fields.Traditional enterprises rely on manual welding,in which the quality of manual welding depends on the technical level and working conditions of the workers,thus the efficiency and quality conformity cannot be guaranteed.Compared with manual welding,automated welding is with high efficiency and consistent quality.The use of automated welding can improve the production efficiency of enterprises and save labor costs.Welding seam tracking technology is an important part for welding control and a necessary condition for welding automation and intelligence.In this thesis,focusing on the research of welding seam tracking technology,we design a welding seam positioning and tracking system based on machine vision.The system designed in this thesis adopts passive vision and uses monocular sensor to collect images.Owing to the monocular vision cannot effectively obtain depth information,this thesis adds an initial height guidance module to the system and realizes the automatic height adjustment function of the camera and welding gun.For the initial height guidance module,we propose proposes a method for evaluating the sharpness of the image without references.This method is based on human visual system,whose evaluation results are in good agreement with the human eye’s evaluation results.It can establish the relationship between the image acquisition height and the sharpness to help the system obtain a suitable initial height.The traditional tracking systems,with weak anti-interference ability and poor segmentation quality,rely on artificial features for image segmentation of the welding seam.This thesis adopts a deep learning method to segment the image and proposes a segmentation network for the welding image by adjusting the loss function and adding attention modules on the ENet.The network can overcome the interferences such as arcs and spatters,and the extraction of the welding seam trajectory is accurate.The system applies this network for segmentation of the welding seam and extracts the centerline of the seam based on the segmentation results.We finally build the welding seam positioning and tracking system and design a graphical interface.Users can observe the real-time tracking results,view the system status and set parameters in the interface.The results from testing the functional modules of the system show that the system has high efficiency and stability and can track welding seam in real time.