基于改进Otsu算法的TFT-LCD点缺陷自动光学检测系统  

Automatic optical detection system for TFT-LCD spot-type defect based on improved Otsu algorithm

作　　者：郭波 管菊花[2] 黄志开[1] GUO Bo;GUAN Ju-hua;HUANG Zhi-kai(Jiangxi Province Key Laboratory of Precision Drive&Control, Nanchang Institute of Technology,Nanchang 330099,China;Jiangxi Institute of Mechatronic Technology,Nanchang 330014,China)

机构地区：[1]南昌工程学院江西省精密驱动与控制重点实验室,江西南昌330099 [2]江西机电职业技术学院,江西南昌330013

出　　处：《液晶与显示》2018年第3期221-227,共7页Chinese Journal of Liquid Crystals and Displays

基　　金：国家自然科学基金(No.61472173);江西省科技厅重点研发计划(No.20151BBE50083)~~

摘　　要：针对TFT-LCD点缺陷自动光学检测时,缺陷与背景对比度较低难以用传统阈值分割算法处理的难题,提出一种改进的Otsu算法,并构建了TFT-LCD点缺陷自动光学检测系统。首先,通过Gabor滤波去除了纹理背景的影响。然后,利用威布尔函数形态参数分段取值时,其分布函数呈现的不同分布特性,改进了传统Otsu阈值提取函数。最后,进行了离线测试试验和在线测试试验。试验表明,改进的Otsu算法在点缺陷与背景对比度较低的情况下分割效果优于传统Otsu算法。将该算法移植到TFT-LCD点缺陷自动光学检测硬件平台上进行在线测试,正确检测率可达到94%,单个样本最短检测时间可缩短至150ms。降低了TFT-LCD人工检测点缺陷的工作量和劳动强度。In order to solve the problem that the traditional threshold segmentation algorithm is difficult to detect the spot-type defect from low contrast background,an improved Otsu algorithm is proposed.Based on this,a TFT-LCD spot-type defect automatic optical detection system is constructed.First,the effect of texture background is removed by Gabor filtering.Then,Using the different distribution characteristics of Weibull distribution function when the shape parameter of the Weibull function is segmented,the threshold extraction function of traditional Otsu is improved.Finally,offline test and online test are carried out.The experiment shows that the improved Otsu algorithm can produce better detection result than the traditional Otsu algorithm in the case of low contrast between defects and background.The algorithm is transplanted to the TFT-LCD spot-type defect automatic optical detection hardware platform to test.The correct detection rate can reach 94%and the detection time can be shortened to 150 ms.It reduces the workload and labor intensity of TFT-LCD manual spot-type defect detection.

关 键 词：点缺陷检测 TFT-LCD 改进的Otsu算法 自动光学检测系统 

分 类 号：TP278[自动化与计算机技术—检测技术与自动化装置]