Research On Automatic Seam Tracking Method And System Based On Laser Vision Sensor

With the continuous development of offshore resources,the demand for marine equipment such as ships is becoming increasingly wide,and the requirements for welding efficiency and welding quality are rapidly increasing.Welding process is the key to the manufacturing quality of offshore equipment,but its automation level restricts the development of the entire industry.TKY jacket joints are mostly manufactured by traditional manual welding,requiring the highest level of 6GR welders,and the consistency and stability of welding quality still need to be improved.Although the teaching online robot has improved the welding efficiency to a certain extent,it cannot fully adapt to the changes of actual working conditions such as clamping error,welding thermal deformation,etc.,and the welding accuracy and quality are difficult to guarantee.Therefore,the research on automatic seam tracking method and system based on laser vision sensing can improve the level of welding automation,and has important practical significance for actual production.In this paper,the overall hardware system of weld seam tracking is designed and built,and the three-dimensional mathematical model of structured light vision system is established.Aiming at the strong interference environment such as arc light,splash,dust and smoke in the welding process,a weld seam feature point extraction algorithm based on image processing and nuclear correlation filtering algorithm(KCF)is proposed.The software interface and function of the automatic weld seam tracking system are developed in combination with Qt,Halcon and Matlab,and tested and verified.The main research contents include:(1)Determine the overall scheme of the seam tracking system and complete the physical platform construction.According to the function and main performance indicators,the key components of the laser vision sensor are selected and the mechanical structure is designed.By analyzing the detection principle of the laser vision sensor seam tracking system,the sensing mode of laser oblique beam and camera vertical reception is determined;The distance between the sensor and the welding gun is 110mm;The distance between the sensor and the workpiece is115 mm and other key parameters.(2)The mathematical model of laser vision 3D measurement with structured light as active light is established.In order to obtain the mathematical model of the laser plane in the robot base coordinate system,the principles of camera intrinsics and extrinsics parameter calibration,structured light plane calibration,and hand-eye calibration are analyzed respectively,and the corresponding relationship between the image coordinates of the points on the laser line and the robot base coordinate system is deduced.The calibration experiment is completed to realize the restoration of the 3D coordinates of the weld feature points.(3)According to the existing mode of noise in the image collected before welding,the welding spot extraction algorithm of the pre-welding structured light weld image is studied,and a real-time welding image preprocessing process and weld feature point extraction algorithm are developed.The center line of the laser stripe is extracted by image filtering,threshold segmentation,ROI extraction,skeletonization and morphological modification of the structured light weld image,and the straight line of the center line of the laser stripe is detected by Hough transform,The pixel coordinates of solder joints are located.(4)Aiming at the problem that the image processing method is easy to be disturbed by arc light,spatter and other noises during the welding process and the weld seam tracking fails,this study uses the position information of the weld groove feature points obtained from the prewelding image processing,uses the kernel correlation filter(KCF)algorithm to locate the weld seam center in real time during the welding process,trains the classifier through samples,and uses the kernel function to detect the position with the largest response value as the target position,Finally,the appearance model is updated online.Combined with Qt,Matlab and Halcon computer vision library,the upper computer software development platform of automatic seam tracking system is built.Finally,the reliability of the algorithm and system is verified by experiments.