Double Line Laser Vision Sensing Seam Tracking Image Processing

This paper studies weld seam tracking based on double line laser vision sensor, including camera calibration, image processing, recognition and location of the initial welding position, and oblique line weld seam tracking. The weld seam tracking based on structured light vision sensor gained wide attention recent years. As a result of lens distortion, the images collected by CCD would be distorted, via camera calibration, the interior, exterior parameters and distortion coefficient of the camera could be obtained, then proceed to undistort images. Images captured during the welding process interfered by arc light, splash and smoke dust. In order to obtain the accurate weld deviation, it has to process the seam image, thus improving the accuracy of seam feature extraction. Recognition and location of the initial welding position is one of the key technologies of intelligent welding, the author used the seam imformation obtained by double line laser vision sensor, located welding torch to the initial welding positon. Based on that, the tracking of oblique line seam achieved favorable result. The idiographic research fall into four parts as follows.Firstly, this paper make use of seam tracking system consist of three-axies robot, welding machine, welding torch, industrial computer, double line laser vision sensor, image acquisition card and so on. Considered the adverse influences caused by lens distortion, the author undistorted seam images by using the interior parameter and distortion coefficient, which is obtained by camera calibration using Zhang's calibration method. The experimental result reveals that image undistortion works well.Subsequently, the double line laser vision sensing seam tracking image has been processed. This paper adapts image processing method include dynamic ROI extraction, gray mophology open operation, image subtraction, LOG filter, isolated blocks remover, thinning, deburring and so on. Then extracted the ROI of deburring image, and used least square method to fit the straight line. Therefore, the feature imformation of seam was extracted accurately. The experiment results show that the algorithm possesses good realtime and strong anti-jamming capacity.This paper builds recognition models of welding torch altitude, seam transverse deviation and trend. On the basis of sensor calibration, we studied recognition and location of the initial welding position, increased the autonomy of seam tracking.Finally, under the VC++6.0 development enviroument, with the combination of OpenCV, compiled the seam tracking software. It mainly contains image acqusition, image processing and initial positon recognition. By means of welding torch altitude recognition and oblique line seam tracking experiment, the deviation was controlled under 0.5mm. The results of weld seam tracking experiment show that the algorithm proposed possesses good realtime, strong anti-jamming capacity and reliable tracking.