基于迁移学习的TFT-LCD缺陷智能识别系统研制与应用

发布时间：2018-03-07 06:23

本文选题：TFT-LCD　切入点：缺陷　出处：《华中科技大学》2016年硕士论文　论文类型：学位论文

【摘要】：近年来,薄膜晶体管液晶显示器(Thin Film Transistor-Liquid Crystal Display,TFT-LCD)已经成为最主流的显示设备,并且朝着高分辨率、大尺寸、轻薄化的方向发展。TFT-LCD制造工艺越来越复杂,为保证产品质量,屏幕缺陷智能识别系统在生产制造中的地位日益凸显。本文针对Cell制程阶段的TFT-LCD缺陷测试的实际需求研制了一套缺陷智能识别系统。具体研究内容主要包括:依据TFT-LCD缺陷的测试需求,设计了缺陷智能识别视觉系统的总体方案。针对Cell制程阶段TFT-LCD缺陷的特点,完成了视觉系统的硬件设计和选型,保证了视觉系统的高精度、高对比度图像采集和高性能处理。针对采集的图像存在不均匀的现象,对视觉系统各模块进行了光学建模,分析其中影响图像不均匀的因素,并对其优化,改善了图像不均匀的现象。针对现有缺陷分类识别算法识别种类少、精度不够高并且处理过程耗时,无法完成在线缺陷识别的问题,提出了一种基于迁移学习的在线序列识别算法(OSC-TL),实现了Mura缺陷的在线识别。针对TFT-LCD缺陷测试设备的运行要求,设计和实现了以视觉处理为核心的软件系统,主要包括软件总体架构设计、接口设计以及人机交互界面设计。本文研制的TFT-LCD缺陷智能识别系统在实际生产中测试,缺陷识别率达到了93%,识别时间约为30毫秒,验证了该系统的有效性,达到了TFT-LCD缺陷在线识别的要求。
[Abstract]:In recent years, thin Film Transistor-Liquid Crystal display TFT-LCDs have become the most popular display devices, and the manufacturing process of TFT-LCD has become more and more complex in the direction of high resolution, large size and thin. The status of intelligent screen defect recognition system is becoming more and more important in manufacturing. According to the actual requirement of TFT-LCD defect testing in Cell process, a set of intelligent defect recognition system is developed in this paper. The main research contents are as follows: according to the requirements of TFT-LCD defect testing in Cell process, the main contents of this paper are as follows:. According to the test requirements for TFT-LCD defects, According to the characteristics of TFT-LCD defect in Cell process, the hardware design and selection of vision system is completed, which ensures the high precision of vision system. High contrast image acquisition and high performance processing. Aiming at the uneven phenomenon of the collected image, the optical modeling of each module of the visual system is carried out, and the factors affecting the uneven image are analyzed and optimized. Aiming at the problem that the existing defect classification and recognition algorithms have fewer kinds, less precision and time-consuming processing process, it can not complete the on-line defect recognition. An online sequence recognition algorithm based on migration learning is proposed to realize the on-line recognition of Mura defects. According to the running requirements of TFT-LCD defect testing equipment, a software system based on visual processing is designed and implemented. This paper mainly includes the software architecture design, interface design and human-computer interface design. The TFT-LCD intelligent defect recognition system developed in this paper has been tested in actual production. The defect recognition rate has reached 93%, and the recognition time is about 30 milliseconds. The effectiveness of the system is verified and the requirement of TFT-LCD defect online recognition is achieved.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41;TN873.93

【相似文献】