Research On Surface Defect Detection Of Magnetic Sheet Based On Deep Learning

The detection of the surface defects on magnetic sheet has played an important role in the production efficiency and the cost of production in the production line of the magnetic sheet factory.A variety of machine vision methods has been applied,they are taken to extract features of artificial defects,but because the disk surface has low contrast,wear texture interference and small changes in the brightness and defects of the difficulties,they lead to less accuracy and versatility;Therefore,a robust defect detection system will bring great benefit to the production of magnetic sheet.Because deep learning has the ability to autonomously learn the characteristics,and to express the structure more accurately,this paper proposes a method of detecting the surface defect of magnetic sheet based on deep learning.By training the lightweight Inception-Resnet-v2 deep neural network model,the classification results are obtained and compared with the traditional machine learning methods.This method can not only automatically extract features and classifications,but also realize the automation of magnetic sheet detection,and the accuracy of recognition has also been greatly improved.In the course of the experiment,it is found that the use of the deep learning model to train the training data with a large number of labels is needed,so the cost of manual annotation is very high.In this paper,a sample optimization method based on active learning is proposed.This method is used to improve the training process of the above deep learning.In each iteration of the training process,a large number of information and the diversity of the samples is selected to train,and the maximum accuracy of the classifier can be achieved by using less training samples.This sample optimization method,while taking into account the accuracy of the classifier,greatly reduces the cost of manual labeling,improves the production efficiency,and is in line with the actual industrial production situation.The experimental results show that compared with the traditional machine learning method,the precision of the surface defect detection method based on deep learning can reach 96.7%,and the active learning optimization training process is adopted to reduce the 25% manual labeling cost.