Research On Surface Defect Classification Method Of Steel Plate Based On Convolutional Neural Network Model

Steel plate is an important raw material in aerospace,machinery manufacturing,defense equipment and other fields.Due to production processes and raw materials,various defects appear on the surface of the steel plate.The appearance of defects on the surface of steel plate will reduce the product quality of steel plate,reduce the mechanical properties of steel plate,weaken the corrosion resistance.Therefore,it is of great significance to detect the surface defects of steel plate in real time in the production process to ensure the product quality of steel plate.Under this background,this paper studies the image preprocessing technology of steel plate surface defects and the classification algorithm of steel plate surface defects.The main research contents of this article are as follows:?1?Image pre-processing of steel plate surface defects.The image preprocessing technology in this paper mainly includes image filtering and image enhancement.Three commonly used filtering algorithms are analyzed,and combined with experimental analysis,the mean filtering is selected as the filtering method for the surface defect image of the steel plate.MSRCR algorithm is used to enhance the defect image,improve the quality of the original image,better highlight the image features,and improve the brightness and contrast of the image.?2?Surface defect classification method of steel plate based on VGG-19convolutional neural network model.Firstly,the data set was expanded by using data enhancement algorithm,then the weight parameters of convolution layer and pooling layer in VGG-19 network were shared by transfer learning,and some other layers were added.Finally,the data set was classified by softmax classifier.The experimental results show that VGG-19 network reduces the size of the model,improves the classification accuracy and reduces the loss value compared with other convolutional neural network models.?3?The classification method of steel plate surface defects based on CNN-SVM.The convolutional neural network?CNN?and support vector machine?SVM?fusion algorithms were used to classify the fine steel plate surface defects.Firstly,the convolution layer and pooling layer in CNN were used for feature extraction,then SVM classifier was used for classification,and ₂L regularization algorithm was added in the training process.The experimental results show that the CNN-SVM method has the highest classification accuracy and the shortest test time.