V-shaped Weld Tracking System Based On Active Machine Vision

Weld seam tracking system based on active vision is widely used in many industrial fields.For the problems of complex feature detection algorithm,long detection time and serious spatter during the welding process of V-butt aluminum alloy weld,this thesis researches a weld seam identification tracking method based on active machine vision to complete the weld seam tracking with short time and high accuracy in the welding work environment with many disturbances.First,this thesis describes the composition of the weld seam tracking system and illustrates the workflow of the whole system.The structure and composition principle of the vision sensor is explained;in order to obtain clear images of the weld seam,a suitable camera and laser mounting method is selected,and also suitable components for each part of the system are chosen.Next,the camera calibration,hand-eye calibration and structured light plane calibration of the vision system were studied.Camera internal reference calibration experiments,hand-eye calibration experiments and structured light plane calibration experiments were conducted to obtain the camera internal parameters,hand-eye conversion matrix and structured light plane parameters to achieve the interconversion between 2D image points and3 D spatial points.The pixel distance calibration was also performed to obtain the actual physical dimensions represented by unit pixels in the images.Then,the study of weld feature point extraction of weld images was carried out.The image was pre-processed with ROI extraction,threshold segmentation and other pre-processing to obtain a binary image containing the weld position information,and an improved algorithm based on the connected domain seed filling method was proposed to obtain the laser streak image,which can effectively eliminate the interference of arc light and spatter present in the welding process.For the case of overlapping spatter and weld laser stripes,a spatter removal method is proposed to quickly remove the spatter stripes while greatly ensuring the integrity of the weld laser stripes.After the pre-processing operation,the median of the row coordinates of the two inflection points of the upper and lower laser stripes and the column coordinates of the end point of the lower half laser stripes are taken as the rank coordinates of the feature points,respectively,and the feature points are calculated.It is verified that the algorithm still has good accuracy and real-time performance under the interference such as arc light and sputtering in welding work.In order to avoid vibrations occurring during the tracking process of excessive deflection,a fuzzy controller is designed and Simulink is used to design simulation experiments to verify the effectiveness of the controller.Finally,according to the parameters that need to be monitored and adjusted in the experiment,the weld tracking software is designed,which can control the start and end of the system.Through the weld seam identification and tracking experimental platform,the weld seam tracking experiments are conducted at a speed of 45cm/min for linear V-shaped aluminum alloy specimens.The test results show that the absolute error of tracking is within0.3mm,and 10-13 weld seam images can be processed per second,which proves that the weld seam identification and tracking method proposed in this thesis meets the requirements of high accuracy and real-time in industry.