A Mobile Phone Screen Cover Glass Defect Detection MODEL Based on Small Samples Learning

Mobile phone screen cover glass (MPSCG) defect detection is an important part to ensure the quality of mobile phone products. In view of the difficulty of obtaining a large number of defect samples on the industrial production line, this paper designs a MPSCG defect detection model suitable for small samples learning. Because of the high resolution of MPSCG original image, the pre-processing software is designed to automatically segment the original high-resolution image into a series of sub-images. Furthermore, the Deep Convolutional Generative Adversarial Networks (DCGAN) network is designed to automatically extract and fuse the defect features to augment and generate defect samples. Then, based on the augmented data set, we improve and train the detection model of Faster R-CNN. The detection model achieved a very better detection result, which solved the problem that the number of defect samples in the industry is small and the deep learning requires a large number of samples. The experimental results demonstrate the effectiveness and feasibility of DCGAN combined with Faster R-CNN for the defect detection of MPSCG.