Research On Laser Welding Seam Tracking System Based On Active Vision

Robotic intelligent welding has the advantages of stable production quality and high efficiency compared to traditional welding,which is a major development trend of future welding processing and manufacturing technology.Laser welding has the advantages of high energy density and high welding speed,which is one of the important aspects of modern welding technology.Laser welding because of the laser focus spot diameter is small,narrow weld seam,pre-processing,assembly accuracy and welding process positioning accuracy requirements are high,if the welding process deviation in positioning,easy to cause welding defects.Therefore,real time extraction and tracking control of the weld spot in the laser welding process is the key to ensure the quality of the weld,and is now an urgent problem to be solved.For the problems of manual alignment deviation of welding guidance,inconspicuous light bar distortion of narrow weld seam and complex and variable welding process,this thesis investigates the initial laser welding position guidance and real-time seam tracking control system based on active vision guidance.The thesis designs and establishes an active vision-based initial weld guidance and tracking system,designs and customizes the structure of the sensor body through 3D modeling,designs the hardware composition of the vision sensor and selects the required devices,realizes real-time image acquisition and hardware denoising;according to the built vision system,analyzes the vision model,calibrates the camera parameters and light plane parameters of the sensor,and realizes the mapping of weld seam feature points from pixel2 D coordinates to real-world 3D coordinates.In the initial guidance stage of welding,the initial point identification method based on edge template matching is proposed for the phenomenon of poor manual alignment accuracy of laser welding guidance.The method is divided into two steps,with edge-based template matching followed by feature extraction via Harris corner point detection combined with sub-pixel detection.For the problem of pseudo-edges caused by noise in edge detection,a distance denoising method based on Hough straight line is proposed,which can effectively remove residual pseudo-edges.To increase the robustness of the algorithm,the proposed initial point selection rule can filter the correct initial point of the weld.The comparison shows that the proposed method can operate effectively under both uneven and insufficient illumination conditions.Weld seam feature image processing study for narrow gap butt welds.A dynamic ROI region extraction method is proposed to reduce the noise impact during the welding process and to improve the system operation speed.At the same time,in order to realize the extraction of weld feature points,the techniques of pre-processing,image segmentation,morphological filtering,weld straight line and laser stripe centerline are systematically compared,studied and analyzed in depth.For the phenomenon that laser streak features are not significant in narrow gaps,the intersection of the fitted light bar straight line and the weld straight line is used instead of the weld feature points.This thesis builds a pilot system for laser welding seam initial position guidance and seam tracking based on active vision guidance.In order to verify the effectiveness of the algorithm,a reasonable experimental scheme is designed.Among them,multiple sets of tests are conducted for weld guidance,and the results show that the welding initial point guidance accuracy is less than 0.2mm in each axial direction,which meets the welding initial guidance requirements.Finally,the real-time tracking control test was conducted for several groups of narrow gap butt welds,and the tracking accuracies obtained for each group were 0.12 mm,0.167 mm and 0.186 mm,respectively.the accuracy test results show that the weld identification and tracking control system studied in this thesis meets the accuracy and real-time requirements in robot production.