TFTLCDMura缺陷机器视觉检测方法毕昕.pdf

《TFTLCDMura缺陷机器视觉检测方法毕昕.pdf》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、第46卷第12期机械工程学报Vol.46No.122010年6月JOURNALOFMECHANICALENGINEERINGJun.2010DOI：10.3901/JME.2010.12.013*TFT-LCDMura缺陷机器视觉检测方法毕昕丁汉(上海交通大学机械与动力工程学院上海200240)摘要：针对液晶显示器(Liquidcrystaldisplay，LCD)制程中Mura缺陷检测的重要性和人工检测的弊端，研究TFT-LCDMura缺陷的机器视觉自动检测方法。基于国际半导体设备与材料组织(SemiconductorEquipme

2、ntandMaterialsInternational，SEMI)标准中Mura缺陷的测量规范和LCD视觉检测试验平台，针对Mura缺陷边缘模糊、对比度低、图像中存在重复纹理背景和整体的亮度不均匀等特点，分别研究基于实值Gabor小波滤波的纹理背景抑制方法、基于同态变换和独立分量分析的亮度不均匀校正方法、基于主动轮廓模型和水平集方法的缺陷分割以及基于SEMI标准的缺陷量化方法，综合几个方面的研究，建立Mura缺陷自动检测流程。检测试验证明，所提出方法能较好地抑制纹理背景、校正背景亮度不均匀和莫尔条纹，准确的分割缺陷并进行量化评定。该方

3、法适用于Mura缺陷的自动检测，检测方法与人的视觉特性相似，具有较好的鲁棒性。对于50个带有Mura缺陷的LCD样本，有48个样本被成功检测。关键词：Mura缺陷Gabor滤波独立分量分析主动轮廓模型水平集中图分类号：TP274.3MachineVisionInspectionMethodofMuraDefectforTFT-LCDBIXinDINGHan(SchoolofMechanicalEngineering,ShanghaiJiaoTongUniversity,Shanghai200240)Abstract：Theautoma

4、ticmachinevisioninspectionwayisstudiedfortheMuradefectofTFT-LCD,aimingattheimportanceofdefectinspectionandtheshortcomingofmanualinspectioninliquidcrystaldisplay(LCD)process.Muraislocallightnessvariationwithlowcontrast,blurrycontour,unevenbrightnessandtexturedbackground.

5、BasedontheSemiconductorEquipmentandMaterialsInternational(SEMI)standardforMuraandtheLCDvisioninspectionplatform,theinspectionalgorithmsareresearched,includingthetexturedbackgroundsuppressionmethodusingrealGaborfiltering,theadjustmentofbrightnessunevennesswithhomomorphic

6、transformandindependentcomponentanalysis,theMurasegmentationusingactivecontourmodelandlevelsetmethodandtheMuraquantificationbasedonSEMIstandard.Theautomaticinspectionprocessissetupbysynthesizingtheresearchesproposedabove.Theinspectionexperimentsshowthattheproposedmethod

7、cansuppressthetexturedbackground,eliminatetheunevennessandmoirefringeinbackground,accuratelysegmentdefectsandcarryoutquantitativeevaluation.TheproposedmethodisapplicabletoautomaticinspectionofMuradefectwithgoodrobustnessandsimilartovisioncharacteristicofhumaneyes.And,fo

8、r50LCDsamples,48samplesareinspectedaccurately.Keywords：MuradefectGaborfilteringIndependentcomponentanalysisAct