CNN-Based Surface Defect Detection of Smartphone Protective Screen

Abstract

With the advancement of the intelligent era of Industry 4.0, in order to solve the problem of low-efficiency and high miss-detection rate using manual visual inspection for the detection of surface defects on smartphone protective screens, surface defects on smartphone protective screens based on convolutional neural networks are introduced as the detection method. This detection method exhibits excellent feature extraction capabilities and powerful target classification performance during the preprocessing, model design, model training, and detection of smartphone protective screens images. Experimental results show that the detection method can achieve accurate detection of defects such as punctures, bright spots, and scratches on the surface of the smartphone protective screens. The detection verification rate is as high as 98%, and the accuracy is high, which meets the actual needs of enterprises.