Research On Weld Seam Recognition And Tracking System Based On Laser Vision

Welding is widely used in industrial production.Because of low production efficiency and difficulty in guaranteeing welding quality,traditional manual welding is gradually being replaced by automated welding technology.Therefore,this thesis studies the welding seam recognition and tracking system based on laser vision.Based on the active visual recognition of the weld seam characteristics,the welding robot is guided to complete the welding seam tracking.First,the laser vision system and its calibration method are studied.The imaging model of the camera is analyzed,the digital model of the camera is established,and the parameter matrix of the camera is obtained through calibration experiments.A practical calibration method for the Eye-in-Hand model is used to obtain the relative position relationship between the camera and the end of the robot,and determine the conversion relationship between the camera coordinates and the space coordinates.Secondly,in order to extract the feature information of the weld image,a preprocessing flow of the weld image is established by analyzing different image processing methods.The image is grayscaled based on gamma transformation,and adaptive median filtering is used for image filtering.A threshold segmentation method optimized based on the otsu method is proposed to threshold the image,and finally clear laser stripes are extracted.Then,the method of extracting feature points of weld image is studied.The appropriate method is determined to extract the centerline of the laser stripe by analyzing and comparing different image refinement algorithms.According to the similarity of the laser stripes of the adjacent weld images,by limiting the parameter search range,using the improved Hough transform to fit the straight lines on both sides of the feature points,the weld feature points are obtained,and the image processing efficiency is improved.Finally,the design of the experimental system is completed according to the content of this thsis,and the performance of the system is verified through multiple sets of weld recognition and tracking experiments.The results of multiple sets of weld tracking experiments show that the average tracking error of the system in the Y-axis direction is 0.47mm,and the average tracking error in the Z-axis direction is 0.75mm,which can meet the actual welding needs of fillet welds.