Research On Intelligent Algorithms In The Classification Of Strip Steel Defect Images

With the development of productive forces and the improvement of living standards, people put forward higher requirements on the quality of iron and steel products. As the main products of steel, steel strip has become an indispensable necessity in every industry. However in the production process, due to the influence of a variety of reasons, so that its surface is inevitably there will be scabs, cracks, cracks, scratches, inclusion and so on different types of defects, which seriously affects the performance of the product. Automatic, accurate and real-time detection of defective products is a long-term and challenging problem in the steel industry.How to classify and detect the defects in the steel production line is not only the problem of strip defect image, but also can be applied to the classification and detection of gray, black and white and even color images of the rest of the industry. In order to improve the accuracy of the strip defect image classification, which is a more complex optimization problem (similar to NP difficult problem), this is a more complex optimization problems. Because of its unique intelligence, intelligent optimization algorithm can find better solutions in a short time. The algorithm of support vector machine is studied by the theory of asymptotic performance. Several intelligent optimization algorithms and support vector machine are integrated to design a hybrid algorithm. The major contributions of this thesis are as follows:1. As for the strip defect image classification problem, it's necessary to pretreat thestrip defect image. We put forward two kinds of binarization method, based on the general genetic algorithm of binarization strip defect image processing algorithm, as for the light of the strip defect image, and put forward the genetic algorithm based on top hat transform binarization strip defect image processing algorithm. Experimental results show that these two methods are better than the traditional OSTU method in terms of time consumption and the image effect of binarization and the latter algorithm is better than the former algorithm.2. There are a lot of research on classification, however, which is more suitable for the classification in strip defect image, and it is not related to it. This thesis analyzes and compares the classification method based on the strip defect image classification. The experimental results show that the classification effect of SVM is better than that of BP neural network, which is the basis for the further optimization of the following chapters. The results show that the classification effect of SVM is better than that of BP neural network on the whole.3. A classification algorithm based on the genetic algorithm and support vector machine (SVM) is proposed:the feature vector, the kernel function type, the parameters and penalty factor of the strip defect image are combined together. Experimental results show that the algorithm has good performance in time and classification accuracy, which can meet the real-time and accuracy of the classification of strip defects.4. For support vector machine parameters optimization, the classification algorithm of particle swarm algorithm and support vector machine is designed, based on the feature vector of strip defect image, kernel function type, parameter of support vector machine, and the penalty factor. The algorithm can be used to improve the model parameters, kernel function and strip defect image.5. The algorithm is proposed based on integration of steel strip defect image of artificial bee colony algorithm and support vector machine (SVM) classification algorithm:the steel strip defect image feature vectors, support vector machine of the type of kernel function, support vector machine parameter and penalty factor combined together to form a composite location of the nectar source code. Through the optimization of the artificial bee colony algorithm, it can find the optimal location of the nectar source, i.e., obtaining the optimal solution. Experimental results show that the algorithm has excellent performance in time and classification accuracy, and can meet the real-time and accuracy of strip defect image classification.6. The convergence of genetic algorithm and support vector machine (SVM) is proposed to improve the convergence of the algorithm, based on the combination of quantum genetic algorithm and support vector machine, which combines the feature vector of strip defect image, support vector machine, support vector machine, and penalty factor. The experimental results show that the algorithm has the same effect on the classification accuracy of the algorithm and the genetic algorithm and the support vector machine, but the time consumption of the algorithm is reduced.