液晶屏（liquid crystal display, LCD）和有机发光半导体（organic light-emitting diode, OLED）屏的制造工艺复杂，其生产过程的每个阶段会不可避免地引入各种缺陷，影响产品的视觉效果及用户体验，甚至出现严重的质量问题。实现快速且精确的缺陷检测是提高产品质量和生产效率的重要手段。本文综述了近20年来基于机器视觉的液晶屏/OLED屏缺陷检测方法的研究进展。首先给出了液晶屏/OLED屏表面缺陷的定义、分类及其产生的原因和缺陷的量化指标；指出了基于视觉的液晶屏/OLED屏表面缺陷检测的难点。然后重点阐述了基于图像处理的缺陷检测方法，包括介绍图像去噪和图像亮度矫正的图像预处理过程；考虑到所采集的液晶屏/OLED屏图像存在纹理背景干扰，对重复性纹理背景消除和背景抑制法进行分析；针对Mura缺陷边缘模糊等特点，总结改进的缺陷分割方法；阐述提取图像特征并使用支持向量机、支持向量数据描述和随机森林算法等基于特征识别的缺陷检测方法。接着综述了基于深度学习的缺陷检测方法，根据产线不同时期的样本数量分别总结了无监督学习、缺陷样本生成、迁移学习和监督学习的方法，其中无监督学习从基于生成对抗网络和自编码器两个方面进行阐述。随后梳理了通用纹理表面缺陷数据集和模型性能的评价指标。最后针对目前液晶屏/OLED屏缺陷检测方法存在的问题，对未来进一步的研究方向进行了展望。
Vision-based LCD/OLED defect detection methods: a critical summary
Lin Siyuan, Wu Yiquan(Nanjing University of Aeronautics and Astronautics)
The new display industry is an important foundation for strategic emerging information industries. Under the active guidance and continuous investment of various national industrial policies, China"s new display industry has developed rapidly and has become one of the most dynamic industries. The industry scale accounts for up to 40% of the global display industry, ranking first in the world. Under the background of the current digital information age, the demand for consumer electronics such as smart phones, tablets, computers, displays and televisions in various occasions is constantly rising, resulting in the global demand for LCD (liquid crystal display) and OLED (organic light emitting diode) screens and other display panels showing a rising trend year by year. The manufacturing process of LCD and OLED is complex, and every stage of the production process will inevitably produce various defects, affecting the visual effect and user experience, and even leading to serious quality problems. Fast and accurate defect detection is an important means to improve product quality and production efficiency, So the defect detection in the production process of LCD and OLED is very necessary. This article reviews the research progress of defect detection methods for LCD/ OLED based on machine vision in the past 20 years, hoping it can provide valuable reference. Firstly, the structure and manufacturing process of commonly used TFT-LCD and OLED are given. The defects on the surface of the LCD/ OLED are classified according to the causes of defects, defect size and defect shape, the definition of the defects are given and the causes of the defects are briefly described. The quantitative indicators of defects SEMU and DSEMU are given. The difficulties of surface defect detection of LCD/ OLED screen based on machine vision are explained; Then this paper focuses on the defect detection methods based on image processing. In actual production, the images to be detected are captured by industrial cameras, and their images are easily affected by noise and light source. Firstly, the image preprocessing process of image denoising and image brightness correction is introduced. Then, due to the texture background of the collected LCD/OLED images, it is necessary to eliminate the interference of texture background before segmentation and localization of defects. The repetitive texture background elimination is elaborated, and the defect detection method based on background suppression method is introduced from the three methods of polynomial fitting, discrete cosine transform and statistical analysis, and the measurement standards of background suppression are given. Because Mura defects are characterized by low contrast, blurred edges and irregular shape, traditional edge detection and threshold segmentation methods are not suitable for Mura defects segmentation and it is difficult to achieve reliable detection of Mura defects. So improved defect segmentation methods are introduced in three sections: threshold segmentation and cluster segmentation, active contour model-based method, edge detection and shape detection, and the evaluation indexes of defect segmentation effects are given. Features are the features that can be distinguished from other types of images, extracting local or global features of images and classifying them is also one of the methods of defect detection. The defect detection methods based on feature recognition, which extract image features and use traditional machine learning such as support vector machine, support vector data description, fuzzy pattern recognition, random forest and so on, are explained. No matter the traditional feature extraction method or the classical background reconstruction method, the missing rate of low contrast defect and small area defect is still very high. The traditional defect detection is carried out in multiple steps, which leads to the loss of defect information, resulting in low contrast defect missing and restricting the detection accuracy. The poor expression ability of manually extracted features also leads to the limitation of detection accuracy. In recent years, deep learning has achieved great success in object detection, which can achieve fast and accurate target identification and detection, and more and more scholars have applied it to the defect detection of LCD/ OLED. This paper reviews the defect detection methods based on deep learning. According to the number of samples in different periods of production line, unsupervised learning and supervised learning, as well as transfer learning and defect sample generation methods are summarized respectively; Unsupervised learning based on deep learning includes generative adversarial network and autoencoder, they learn the defect-free samples, reconstruct the defective image in the test, and obtain the residual image for defect detection. Supervised learning requires a large number of defect samples, and to overcome the problems of texture background interference, different defect sizes and uneven samples. At present, there is no public dataset based on display defects. This paper summarizes a series of general texture surface defect data sets that can be used for texture-based background defect detection, which can be used for transfer learning and algorithm universality verification and evaluation indicators of model performance are introduced; Finally, the problems existing in the current LCD/ OLED defect detection methods are pointed out, due to the detection difficulties caused by the characteristics of Mura defects such as low contrast and blurred edges, complex background problems are still unavoidable problems in the detection process, and there are also problems of limited datasets, real-time algorithm problems and so on. The future research direction is prospected, it is pointed out that dataset expansion, achieving sample equalization, enhancing algorithm generality, transferable algorithm, deep learning model acceleration and curved screen defect detection are important research directions in the future, and this may greatly promote the application of machine vision technology in LCD/ OLED defect detection.