| The recognition and classification of surface texture defects of metal work-pieces is a hot issue in the field of industrial automation in the 21st century.Defects such as scratches,spots and pits on the surface of metal work-pieces,it may cause problems such as shortened service life of the assembled machine and potential safety hazards.In order to ensure the quality of products,in this paper,the methods of extracting and classifying surface texture defects of metal work-pieces are studied.In this paper,based on the study of gray co-occurrence matrix algorithm to extract defect features,the relevant mechanism of texture defects is analyzed.Aiming at the gray-scale jump information of texture defect in image,a gray-gradient co-occurrence matrix algorithm is proposed.In this method,the gradient information of the texture in the image is introduced into the grayscale co-occurrence matrix so that it can be reflected from the gradient direction for the directional texture.Experimental results show that the defect features extracted based on the gray grade-gradient symbiosis matrix algorithm can better retain texture edges,ridges or other sharp texture information.In order to improve the classification accuracy of texture defects.In this paper,particle swarm optimization based on linear decreasing weight method is proposed to improve RBF neural network classification algorithm.This method uses particle swarm optimization(PSO)algorithm to optimize x_i,sigma variance and output weight of RBF neural network,and train the network.Thus,the problems of slow convergence and divergence of standard RBF neural network are improved,and the accuracy of defect classification is improved.The experimental results show that the PSO improved RBF neural network improves the classification accuracy of defects compared with the standard RBF neural network and BP neural network classification method.The average classification accuracy of the three defect features was as high as 94%. |
PDF Full Text Request