Design And Implementation Of Surface Defects Detection System For Steel Plate Based On Improved Dagsvm

In recent years, the wide application in steel plane field improves some requirements, especially the quality. In process of the actual production, defects often appear in the surface, interior and edge of steel plate. Among of them, the surface one is more common are complex, which will lead that it is difficult for most of defects detection systems to accurately identify the defect information in actual process. Above systems not only cause economic losses, but also do not conducive to the sustainable development of production. Therefore, it is essential to improve the system based on the theory and practice to improve the accuracy rate in terms of current surface defect detection system for steel plate.In the surface defects detection system for steel plate, the foundation of defect detection is image acquisition. High quality images can not only simplify the complexity of the defect detection algorithm, but also speed up the defect detection. It is important to improve the detection efficiency of the system. In order to obtain high quality defect images of steel plate, this paper improves the image acquisition module in the current surface detection system for steel plate. Firstly, the weakness of lighting module is analyzed in this paper. And we adopt laser lighting module to overcome the problems of poor image quality, which is leaded by uneven illumination and unreasonable illumination angle. Then, we also analyze and compare the current array CCD(Charge Coupled Device) and linear CCD imaging module. According to the selection requirement about linear CCD camera of the system, the suitable linear CCD camera is selected to obtain high quality image of steel plate.Due to the problems of low speed and low detection accuracy in the current system, this paper designed designs a method of surface defects detection for steel plate based on improved DAGSVM(Directed Acyclic Graph Support Vector Machine). Firstly, we analyze the characteristics of the image of defect steel plate in order to propose the preliminary classification method, which can divide the histogram into several regions. Based on the method, we can enhance image to obtain accurate defect target. Then, images can be uniformly processed to improve the speed of defect detection. Due to the more complex defect feature of steel plate and unclear boundaries between different types of defects, we integrate fuzzy thought into the design of the classifier. To obtain optimal clustering center, we apply the characteristics of global optimization in QPSO(Quantum-behaved Particle Swarm Optimization) algorithm to compensate for insufficient FCM(Fuzzy c-Means Clustering) algorithm. And the clustering center is applied to DAGSVM classifier to identify the different defects and obtain accurate detection results. Then, the information of defect details is displayed to operators by the interactive interface.The test image data, which comes from production field, demonstrates that the surface defects detection system for steel plate based on improved DAGSVM can improve the accuracy, reduce the cost of production, improve profit and be beneficial to the long-term development of the enterprise.