电容触摸屏基板ITO电路缺陷自动光学检测关键技术研究

发布时间：2018-08-11 17:48

【摘要】：电容触摸屏(Capacitive Touch Panel,简称CTP)作为目前最主要的人机交互媒介,被广泛应用于平板电脑、智能手机及自动取款机等各类设备中。而在CTP制造中,基板刻蚀ITO电路图案是其中最重要的环节,其是否有缺陷将直接决定CTP的性能。因此,对基板刻蚀后的ITO电路图案进行缺陷检测显得尤为重要。本文以此为目标,对自动光学检测系统中涉及到的关键技术开展研究,主要研究内容与成果如下:(1).基于最大叠加反射系数的高对比度清晰成像系统设计;针对可见光下ITO电路图案几乎透明,无法与基板背景有效地区别开来的问题。本文首先对成像系统进行建模分析。在综合成本与硬件易获性的基础上,构建了基于最大叠加反射系数的高对比度清晰成像系统。该成像系统可使两者(ITO电路图案与基板背景)成像区分度达15.5%。从而达到有效使ITO图案电路凸显区别于基板背景的目的。(2).基于一维线图像的Brenner函数自动对焦;安装误差及人工经验不足势必引起对焦设置不准确的问题,而该问题与设备制造误差一并,又会导致线阵CCD的运动平面与成像平面差超出景深等一系列问题。针对该问题,本文对垂直同轴照明系统进行理论建模分析,得出景深计算公式。然后依据线阵CCD整个扫描过程中对取样点一维线图像的Brenner函数值,自动描绘实际成像平面并规划最佳的相机对焦距离。该系统的实现可以中和由于各种原因引起的图像失焦问题而自动捕捉到最佳对焦平面。(3).基于非负矩阵分解(Nonnegative Matrix Factorization,简称NMF)的大尺寸图像整体配准;由于线阵CCD成像分辨率高的特点,ITO电路图案的成像尺寸巨大。但图像配准是在基于模板的缺陷检测算法中最重要的步骤之一。因此,针对大尺寸图像整体配准效率低的特点,本文研究了一种基于NMF的大尺寸图像配准技术。该方法可以有效的提取配准特征数据,减少配准图像的计算量,有效提升配准时间。(4).广泛适用的缺陷提取算法;本文分别提出并研究了一维线图像比较算法的缺陷提取算法、基于NMF的公差模板缺陷提取算法及陷波模板缺陷提取算法。并对这三种算法进行了横向对比及应用优势分析。同时对算法进行了参数化设计,可以根据不同的成像系统,方便地自动调整检测算法参数,自适应不同种类图案的CTP ITO电路缺陷检测。(5).一种快速的基于缺陷边界变化次数的缺陷分类方法的设计;根据CTP中基板ITO电路图案缺陷的特点,不给整体算法增加负担的基础上,结合一维线图像比较算法的缺陷提取算法与基于NMF的公差模板缺陷提取算法,设计了一种快速的基于缺陷边界变化次数的缺陷分类方法。实验表明该方法快速、准确,可有效对CTP生产过程中的短路、突起、ITO残留污点、开路、针孔、缺口缺陷进行分类。
[Abstract]:Capacitive touch screen (Capacitive Touch Panel, (CTP), as the most important human-computer interaction medium, is widely used in tablet computers, smart phones, automatic teller machines and other devices. In the manufacturing of CTP, the substrate etching ITO circuit pattern is the most important link, whether it has defects will directly determine the performance of CTP. Therefore, it is very important to detect the defects of the ITO circuit pattern etched on the substrate. In this paper, the key technologies involved in the automatic optical detection system are studied. The main research contents and results are as follows: (1). The design of high contrast clear imaging system based on the maximum superposition reflection coefficient is aimed at the problem that the ITO circuit pattern is almost transparent under visible light and can not be effectively distinguished from the background of the substrate. Firstly, the imaging system is modeled and analyzed in this paper. Based on the combination of cost and hardware accessibility, a high contrast and clear imaging system based on the maximum superimposed reflection coefficient is constructed. The imaging system can distinguish the two (ITO circuit pattern and substrate background) up to 15. 5%. In order to achieve the purpose of effectively making the ITO pattern circuit stand out from the background of the substrate. (2). The Brenner function based on one-dimensional line image can focus automatically, the installation error and the lack of manual experience will inevitably lead to the problem of inaccurate focus setting, which is accompanied by the equipment manufacturing error. This will lead to a series of problems such as the difference between the moving plane and the imaging plane of linear CCD exceeds the depth of field and so on. To solve this problem, the vertical coaxial lighting system is modeled and analyzed theoretically, and the depth of field formula is obtained. Then according to the Brenner function value of one dimensional line image of sampling point during the whole scanning process of linear CCD, the actual imaging plane is automatically described and the best camera focus distance is planned. The realization of the system can automatically capture the best focal plane due to the problem of image defocusing caused by various reasons. (3) Large scale image registration based on nonnegative matrix decomposition (Nonnegative Matrix Factorization,) is presented. Because of the high resolution of linear CCD, the image size of ITO circuit pattern is very large. But image registration is one of the most important steps in template-based defect detection algorithm. Therefore, in view of the low efficiency of large scale image registration, a large scale image registration technology based on NMF is studied in this paper. This method can effectively extract the registration feature data, reduce the computation of the registration image, and effectively improve the registration time. (4). In this paper, the defect extraction algorithms of one-dimensional line image comparison algorithm, tolerance template defect extraction algorithm based on NMF and notch template defect extraction algorithm are proposed and studied respectively. The three algorithms are compared horizontally and the advantages of application are analyzed. At the same time, the algorithm is parameterized, which can automatically adjust the parameters of the detection algorithm according to different imaging systems and adapt to the defect detection of CTP ITO circuits with different patterns. (5). The design of a fast defect classification method based on the number of defect boundary changes, according to the characteristics of ITO circuit pattern defects in CTP, does not add burden to the whole algorithm. Combining the defect extraction algorithm of one-dimensional line image comparison algorithm with the tolerance template defect extraction algorithm based on NMF, a fast defect classification method based on the number of defect boundary changes is designed. The experimental results show that the method is fast and accurate, and can effectively classify the short circuit, residual spots, open circuit, pinhole and notch defects in CTP production process.
【学位授予单位】：华南理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TN873

【相似文献】