Real-time Tracking Welding Of Intersecting Line Welds Based On Structured Light Stereo Vision

Pipeline weldments are widely used in modern industrial manufacturing and national pillar industries files.Naturally,the intersecting pipes which is an important part of pipeline weldments is also widely used.However,due to the problems such as difficulty in positioning weld seam and difficulty in planning the welding torch posture when the spatial intersecting pipes were weld,it still relies on manual welding,traditional robot teaching-reproduction and off-line programming and other welding methods which exists many drawbacks.Therefore,the research on the automatic and intelligent welding of the spatial intersecting seam has great scientific research and engineering application value.In this paper,the real-time spatial intersecting seam tracking technology is studied by the combination of monocular structured optical stereo vision sensor and industrial robot.Firstly,aiming at the location problem of intersecting seam,the camera calibration,hand eye calibration and structured laser plane calibration in stereo vision sensor of Eye-in-Hand system were studied.Through the research on the design of vision system and the existing calibration methods,a whole vision system calibration scheme based on the mixed image of "checkerboard & laser stripe" was selected,which reduces the calibration time relatively,and avoids the operation error,random error and the cumulative error between hand eye calibration and structured laser plane calibration in a large extent,and it has high accuracy.Secondly,aiming at the extracting problem of weld feature points from laser stripe image of intersecting seam,the computer graphics and image processing algorithm were studied.In the situation of no arc noise,the welding seam center line characteristic of laser stripe could be obtained by traditional image processing method such as color recognizing in RGB,thresholding,connected area filtering,morphological dilating and thinning,et al.And then,the position of welding feature points could be located by the proposed Triangular Pseudo-Rotation Feature Pixel Detection method.In the situation of arc noise,the traditional image processing method couldn't extract the feature points of image because of the laser stripe contaminated by heavy arc and splash noise.Therefore,an Extending and Adopting Kalman Filter combined with Spatio-Temporal Context algorithm was proposed to solve the problem effectively.Then,the trajectory planning of welding torch and the posture controlling of welding gun were studied when the intersecting seam was weld in real-time tracking.The posture points of laser stripe were extracted by the proposed Expanding Circle and Arc Length method,combining with the extracted weld feature points,the welding gun coordinate frame was established by the proposed three-point principle,and the relation between the robot and intersecting joint was obtained.Under the scheme of multi-segment welding for intersecting seam,the step-size and posture of mobile weld gun were real-time controlled to ensure the welding torch has smoothness and precision during the welding,which could achieve the trajectory planning of welding torch and the posture controlling of welding gun,making the real-time scanning processing and planning weld well.Finally,aiming at the real-time welding problem of the intersecting seam tracking system,the data communication and transmission scheme of cameracomputer-robot and the whole system architecture were studied.By using the socket communication principle of TCP/IP,a "three thread" architecture scheme was designed.Through the micro division of CPU period between threads,the extraction of weld feature points and the data of robot end pose were processed in parallel to ensure the consistency between the robot end pose data and the weld image collected at a certain time during tracking welding,and to ensure the accuracy of real-time positioning and measurement of Eye-in-Hand robotic stereo vision system.Besides,in order to facilitate the operation of the whole intersecting seam tracking system,a Graphic User Interface was developed.Finally,the welding experiment of tracking for intersecting seam was performed and the experimental results were analyzed and verified,which showed that the tracking system is good in real-time,stability,accuracy and adaptability.