Recognition Of The Cold-rolled Strip Plate-profile Based On Computer Vision

The quality control in the production of cold-rolled strip is paid more and more attention to. Traditional contact-plate-shaped detection method is difficult or impossible to meet rapid and accurate detection accuracy, and its cost is high, spare parts expensive, maintenance costs high. Researchers at home and abroad have made many solutions to the shortcomings of contact flatness detection technology. Non-contact detection technology based on computer vision technology has become the direction of development and research hot spots.Relying on the " the iron and steel industry production quality control system based on the new smart calculated theory " (Natural Science Foundation of Chongqing Municipality, item number: 7369) and "Cold-rolled shape control artificial intelligence system" project (horizontal technology, contract number: GK050121), In this paper, ordinary light source and CCD sensor are used to get board-reflective images, image segmentation and pattern classification technology are used to identify patterns defects in the process of rolling cold-rolled strip.In this paper, the key elements are as follows:①Chapter 2 analyses the types and characteristics of flatness defect and proposes the characteristics of the study object.②Chapter 3 presents shape recognition system and module design based on computer vision technology.③In Chapter 4, on the basis of the shape recognition system programme, the image acquisition system, lighting system and the hardware design of image data processing system are completed; the hardware installation and debugging of the entire shape recognition system are realized.④In Chapter 5, on the basis of the completeness of system hardware and the Visual C++ 6.0 as development platform, the software code compilation of the whole shape identification system has been fulfilled such as image acquisition, image pre-processing, image segmentation, identification classification and the results showed. The key issues of the flatness image segmentation and identification classification are solved, segmentation image edge based on image sequence difference method is proposed, flatness defect information is stressed, the smallest error threshold is used to extract image feature information after difference. Compared with the methods based on neural network and morphology, the algorithm turned out to be simple and effective.Running examination shows that the percentage of hits of the system plate recognition arrives at more than 95.3%, which proves that this system has not only had high recognition accuracy, but also met speed requirements of the cold-rolling online identification. In this paper, the study results have a higher theoretical significance and value to promote the application of computer vision technology in the non-contact industrial detection.