Effective Defect Detection Method Based on Bilinear Texture Features for LGPs

L. Hong, X. Wu, D. Zhou and F. Liu

Ieee Access 2021 Vol. 9 Pages 147958-147966

Accession Number: WOS:000716682900001 DOI: 10.1109/ACCESS.2021.3111410

http://dx.doi.org/10.1109/ACCESS.2021.3111410

Automatic defect detection of light guide plates (LGPs) is an important task in the manufacture of liquid crystal displays. During thermo-printing, defects of tag lines on LGPs may occur easily, and these defects are of two categories: bubbles and missing tag lines. These defects lack salient visual attributes, such as edge-based and region-based features, and as such, traditional methods fail to detect them. To address this, we propose a Dense-bilinear convolutional neural network (BCNN), an end-to-end defect detection network, utilizing Dense-blocks [1], Bilinear feature layers [2], and squeeze-and-excitation blocks [3]. Our network exploits fine-grained texture features, which leads to parameter reduction and accuracy enhancement. We validate our network on our LGP dataset containing 5,860 images from three cases: bubbles, tag line existence, and tag line missing. Our network outperforms AlexNet [4], VGG [5], and ResNet [6] on both the public and our LGP datasets with less GPU memory consumption.