Research Of Technology On Surface Defects Detection For Strip Steel Based On Machine Vision

Silicon steel strip is one of the most important materials of motors and transformers, which surface must be insulated. This is why we need smear multi-layer insulation material on it, the quality of the coating directly affects its final production- the performance of motor and transformer. Therefore we must strictly monitor the quality of coating in the producing process of silicon steel strip. If some defects happened, we must adjust the production technology in time to minimize the defect rate of silicon steel strip. Most of the manufacturers in our country are using artificial visual inspection method to detect silicon steel strip’s surface defects, because of the huge production and the fact that the strip width of 1.5 meters per minute up to over forty meters in production line, manual work hardly to guarantee the accuracy and consistency of testing results. In recent years a technology based on image processing to detect defects have been widely applied to various fields of industrial production, it can ensure 24 hours constant detection in production line for the real-time monitoring of silicon steel strip production, provide timely warning processing and other functions, reduce production losses caused by the timeliness and accuracy of detection process and ensure the consistency and reliability of results. Therefore,developing a system that can achieve automatic monitoring of steel strip surface’s quality is of great significance and a high market value.This paper takes the actual demand for a steel manufacturer in Wuxi as background and studies the silicon steel strip defects surveillance system. Our main contributions are as follows:First of all, we have an actual investigation and design the system based on the research results, such as lighting design, industrial camera installation location, image transmission and so on. Then we choose Gigabit Ethernet TXG-13 industrial cameras produced by German company Baumer as image acquisition camera in our system according to the characteristics of the steel strip.Secondly, according to Baumer industrial cameras Baumer-GAPI v1.7 Development Kit, wedevelop an image acquisition system using Visual C ++ 2010, which achieves image’s acquisition and preservation of industrial camera image taking advantage of the Gigabit Ethernet, and various functions of camera’s settings and control based on its full-featured development kit.Following the above basic work, this paper analyzes the characteristics of various defects in silicon steel strip images, studies correlation theory of digital image processing algorithms and designs a fast monitoring algorithm of silicon steel strip defects:First, we conduct the image de-noising, design Butterworth Homomorphic Filter, improve the image boundary portion and the problem of central region’s light imbalance. Then we select different threshold to handle with image’s Binarization, dilation and erosion for that image’s defect points are too light and too dark and achieve significant defect information. Taking advantage of Canny Operator to segment image, we get the good characteristics of defect edge.After that, the image after preprocessed separately projects a shadow in both directions X-axis and Y-axis. We separately difference them and achieve its center location of defect according to derivative different nature. Finally, we design the linear tracking method along the eight directions based on the center of each defect coordinate, obtain the maximum and minimum values of defect boundary coordinates and ascertain the defect boundary information. After analysing and calculating the defect eigenvalue, we can achieve the detection and recognition of defects.