Aluminum Plate Surface Defect Classification Based On SVM

With the widespread application of sheet metal, nonferrous metal industry and steel industry have increasingly high demands for surface quality of sheet metal. Due to various factors, such as production equipment, external environment and so on, sheet metal like aluminum plate, steel belt tends to have many defects during the process of production. The number of defects, types and locations in the sheet metal determined subsequent processing. In today's competitive world, product surface detection technology is an important way of quality control, and it has been widely used to industrial production process. Product surface detection technology can be divided into two categories. One is the artificial visual method which relies on human vision system. This kind of method was applied in the early industrial production or small production plants. There are many disadvantages such as slow speed, low efficiency, subjective test results and depending on the working experience of workers. The other method is intelligent testing method based on machine vision theory. This kind of method has the advantages of high efficiency, fast speed, objective test results, etc. The rapid development of computer application technology, pattern recognition technology and image processing theory, prompt the improvement of the accuracy of this method.Surface defect is one of the key indicators of the quality of aluminum plate in aluminum processing industry. Defect detection and recognition is the core technology of surface quality control and detection system. This paper implements aluminum plate surface defects classification system based on machine vision and it improved the recognition rate of aluminum plate surface defects. According to the characteristics of aluminum plate surface defect image, using image processing technology and pattern recognition technology, combining classifier, this paper achieved the classification of surface defects of aluminum plate, and it mainly contains the following tasks:(1) summarizing the common defects during aluminum plate production;(2) designing and implementation of algorithm of pretreatment of surface defects of aluminum plate;(3) designing and implementation of plate surface defect feature extraction and selection of algorithm;(4) designing and implementation of SVM(support vector machine) classification principle and classifier algorithm. In the classification of surface defects of aluminum plate, the samples of defects are limited, SVM can solve the classification problem of small samples. The innovation of this paper is apply SVM to the classification of aluminum plate surface defects, by changing the kernel function to improve the recognition rate. In this paper, author has done many tests on designing of classifiers, by contrast, the SVM classifier has the highest recognition rate.