Classification of Guide Rail Block by Xception Model

Abstract

Linear guide rail blocks are used in linear slide rail accessories to scrape off oil stains in the rails, installed on the front and rear ends of the slider. They are also used in milling machines, lathes, automated machines, robotic arms, electronic instruments, and so on. At present, the industry relies on manpower to carry out the quality inspection of this rail block which is difficult to standardize. Thus, automatic and digital deep learning inspection technology is introduced for the inspection. To understand the suitability of deep learning techniques applied to the linear guide block inspection process, we adopt the convolutional neural network model architecture and use the Xception model. In model training, the training effect is improved by amplifying the image method and testing many different defects. Through the Xception model, the training accuracy is about 98.7% after 30 epochs, the validation accuracy is about 97.4%, and the test accuracy is about 91.8%.