Charaterization Of Submerged Arc Welding Seam Formation Based On Binocular Stereo Vision

Long-distance oil pipelines need high quality for its welds. The weld has not only a high strength and toughness, but also good corrosion resistance. At present, the surface of pipe is coated by preservative with certain thickness. In this way, the corrosion resistance of base metal and weld surface can be improved. On the other hand the quality of weld formation can directly determine the thickness and effect of preservative. Therefore, it is important to study the technology—automatic detection of weld formation. It not only provides a scientific evaluation for weld formation, but also basic data for embalmed. In this paper, a digital evaluation method is explored based on binocular stereo vision system. The images of submerged arc weld surface gained by the system are used as source of information. The main content is shown as following:(1) A test system of stereo vision based on computer is built to detect the formation of weld. Based on the system, the functions as acquisition, display, storage, processing of the image, parameter extraction etc. can be achieved.(2) The visual system is calibrated in order to create the relationship between the location of image pixel and three-dimensional scene. Then it is corrected to simplified camera model. Internal and external parameters of the two cameras and projection matrix are obtained by calibration based on'Tsai two steps'method. Afterwards, the camera model is corrected according to the transformation of projection matrix. It greatly simplifies the later calculations.(3) The flat information (weld length and width) of weld is characterized by the segmentation of weld image. Firstly, gray-scale normalization and Gaussian filtering method are used in preprocessing of weld image. Secondly, the weld region is segmented by the method of twice dynamic threshold. Finally, binary image of weld that has been segmented is smoothed and filled up by the method of closing operation in morphological processing. The flat contour finally extracted can consistent with the reality.(4) The height information of weld is characterized by stereo vision. Different factors such as similarity measure function, matching window and the similarity threshold, etc. are studied in gray correlation matching method. Based on the study, a method that can get the best algorithm and parameters of stereo matching is explored. Finally, the surface of weld is reconstructed in this way. Reinforcement of each point on the surface of weld can be reflected in the image of three-dimensional reconstruction.(5) A method that can parametrically characterize weld formation is explored. Firstly, the formation parameters of weld width, reinforcement and transition angle are extracted. Secondly, evaluation index of weld formation such asσB(standard deviation of weld width),σc(standard deviation of reinforcement),αL(left transition angle),αR(right transition angle),ψ(weld width ratio) is defined. The calculated results compared with the actual weld shows that the digital indicators extracted can be used as evaluation quantitative indicators of weld formation. Finally, weld formation can be comprehensively evaluated based on weld formation indicators and the current general guidelines.