Method Research Of Welding Seam Tracking Based On Active Vision Sensor

Welding, as a method, play a very important role in the manufacture of modern equipment. As one of the key technology to automatic welding, seam tracking get rapid development, also become an important research content. In this article, structure light of laser-based seam tracking system is studied.In this thesis, through analysis of CCD and CMOS camera characteristics from their working principle and applications, select the CMOS camera to be the tracking sensors. By comparing the camera and laser setup way, let laser in the tilt direction to irradiate, camera in the vertical direction to capture the image.The image pre-processing stage which contains the gray-scale conversion, the filter processing and binarizating is analysised. We use the minimum value of the R, G, B or one value of them to let the gray-scale conversion to come true, and by comparing the template of different shape and size of median filter and mean filter, 5×5 median filter template is selected to filter the image, histogram methods is used to obtain automatic threshold for binarization processing, successfully extracted the structured light image point from the background image. The problems of large quantity data in color image, the difficulties of processing, small regional issues noise points are solved.Image processing steps and algorithms of the stage of getting welding image feature points are analyzed; the second filtering processing, the centerline extraction algorithm and the algorithm for extracting feature points are studied in this stage. In the second filtering processing, the elimination of small regions and Horizontal projection algorithms are used, and then solitary points and large splash noise in the image are flitted. In centerline extraction, the center line extraction after edge detection method, refining processing and the method of by use the max value of horizontal projection are studied, and then the way to get the Feature Point is discussed. Finally, we analysis the feature point extraction method of different shapes of the weld groove, and give each groove the solutions.A corresponding software system is programmed in Microsoft Visual C++ 6.0. In this system, the Program of Image Acquisition, Image Processing and the Communication between the Computer and MCU are prepared. Use the COMS to obtain image, process the image to obtain the welding deviation, then transport the deviation to the MCU, the whole process above is guaranteed in the system.In order to achieve the above mentioned study, an auto-tracking welding device is designed and manufactured in this thesis. A Stepping Motor is used in the device to drive Ball Screw, and then the Torch is driven to achieve real-time tracking. Through the experiments, the device can achieve the tracking request.