The Weld Seam Recognition Of Pipeline NDT System Based On Color Information

With the booming energy industry, the transport of large-sized and distant pipelineshave been popularized. The key factor which affects the transport reliability is thequality of welding pipes. There are common limitations of welding pipes such asventages as well as some situations like impurities, cracks, half-welded, half-alloyed andso on. Thus non-destructive testing (NDT) of Ray is widely used as one of therequisite and effective tools for modern progress because this testing is a kind of earlierdevelopment and mature technology. The advantage of this testing, with the premise ofnon-destroyed, can test the goods’ quality and function with the features of flexibleoperation, production efficiency improved, safety reliability enhanced and so on. TheNDT robot as a new equipment for welded pipe quality testing, the research andapplication have been one of the hot research issues among domestic and foreignscholars. The advanced weld seam recognition technology can achieve a goal ofradiation safety by improving the accurate positioning of the robot system, detecting thewelding quality, improve the safety and economy of the actual operation, avoiding thefollow-up non-destructive testing and reducing direct contact with the radioactivesources, as well as the effects of radiation flaw detection process. In this case theprocessing of welding image is an essential part in the field of welding research, and it isextremely important. Through the acquisition of industrial cameras to the welding image,the image processing system will be followed by the welding image processing to obtainthe corresponding parameters of the welding location, and then it will send them to thecontrol mechanism, to be in charge of the welding real-time flaw detection.The typical way of welded seam identification is based on the original images,considering how the grey levels of pixels change in certain target field, and then usingthe method of image edge testing to find the weld seam. In recent years, at the background of chromatics which has been further studied, people have attached moreimportance to the technology which is based on extracting information of weld seamfrom color images and applied it to various areas, such as welded seam tracing,non-destructive testing of welded seam, visual identification of welded robots and so on.It is helpful to understand images and robot’s position by focusing the places of weldseam out of automation, extracting the information of center lines and identifying andanalyzing them.For color welding image, its color in welding area shows different color features fromthe original material after milling. It seems brownish red due to corrosion and otherreasons in original material. However, it seems cyan as the metal itself after milling. But,in the welding process, affected by the temperature, part of the area which has similarcolor with the original material exists in the welding area after milling. In this area, wecan’t recognize welding by understanding the color information. So when we will wantto directly use the traditional edge detection method or threshold segmentation method,get the position of the welding very hard. In the filming process, the light distribution isuneven, increased the difficulty of the weld we identify.Supported by General Administration of Quality Supervision, Inspection andQuarantine fund （Grant No.201210044-03）, this paper based on the color image,introduce a weld image recognition steps with several ways of gray-scale informationand color information by contrast, focus on the analysis with the oil and gas industrypipe welds image characteristics, introduce the working principle of γ-rays pipe weldsinspection robot system which is used in the image processing. The full text involvescolor weld image vector median filter, color enhancement, contrast to the selected colorspace conversion, mathematical morphology and threshold segmentation. To thecomplex color images color, it includes the uneven illumination phenomenon, so weconsider one method how to extract color information, after that we will segment thetarget and background. In this way, we can get high quality welding area image anddraw the center line of weld. We improve the algorithm, and put forward a new way torecognize the weld. The new recognition algorithm can keep the more weld message, isgood for knowing the weld image. By series of experiments, we proposed the methodeffectively, for the robot system tracking, positioning and ray NDT weld laying a goodfoundation.