Research on Robot Welding Seam Tracking Based on Image Processing Technology
Here was the research on a robot welding seam tracking method based on image processing technol-ogy to solve the problem of low seam tracking accuracy when welding robots face different welding conditions.Using CCD sensors to capture images of robot welding seams,using point operation based image enhancement methods,we could get grayscale transformation of robot welding seam images.Using the SegFormer image semantic segmentation model,a Trans-former encoder and a multi-layer perceptron decoder were used to extract the weld seam tracking path from the grayscale image of the robot welding seam using a recursive cavity self attention mechanism.Using Kalman filtering algorithm to elim-inate non welding points in the weld seam tracking path,updating the welding seam trajectory stack through coordinate sys-tem transformation of robot welding seam images,we could get robot welding seam tracking.The experimental results showed that this method could accurately track welding seams based on the image processing results of robot welding seams,with tracking errors of less than 3mm for arc and S-shaped welds.