液晶屏线路中导电粒子压合的自动光学检测研究  

Detection of conducting particles bonding in the circuit of liquid crystal display

作　　者：陈玉叶[1] 肖可[1] 郭振雄[1] 何俊杰[1] 刘畅[1] 陈松岩[1] CHEN Yu-ye XIAO Ke GUO Zhen-xiong HE Jun-jie LIU Chang CHEN Song-yan(Xiamen University, Physical Science and Tech., Xiamen 361001, China)

机构地区：[1]厦门大学物理科学与技术学院,福建厦门361001

出　　处：《液晶与显示》2017年第7期553-559,共7页Chinese Journal of Liquid Crystals and Displays

基　　金：福建省高校产学合作项目(No.2016H6026)~~

摘　　要：在薄膜晶体管液晶显示器线路检测中,常通过对线路中的导电薄膜粒子的计数和定位实现其导电性的自动检测。为了解决窄边框线路中粒子密度增大带来的粒子重叠问题,提出一种采用微分干涉成像和掩模法结合k均值聚类的算法,在分离出粒子的亮、暗部后,结合图像熵值和粒子的凸性准确分割出粒子。讨论了聚类簇选值的影响,通过不同粒子密度、不同粒子尺寸的样本检验本文算法,并与以往的梯度结合灰度的方法进行对比。结果表明:本文算法在粒子密度较小的区域能达到92.6%的识别率,在粒子密度较大的区域也能达到86%的识别率,分别比梯度加灰度的方法提高了9.9%和42.7%。解决了粒子重叠的问题,并且对光场和成像效果有更好的鲁棒性。By counting and locating anisotropic conductive film（ACF）particles in the circuit of thin film transistor liquid crystal display（TFT-LCD）,it can determine the circuit′s conductivity.In order to solve the overlap problem caused by density increasing of particles in narrow bezel,we put forward a algorithm based on differential interference contrast（DIC）imaging,the algorithm integrates mask method and k-means clustering detection algorithm.After separating particles of bright and shadow,we can effectively segment the particles by judging the entropy of image and the convexity of particles.The value of clustering cluster is discussed,and comparing with the previous method based on gradient and gray level,we test the samples of different particle density and particle size with our proposed algorithm.It indicates that in the case of circuit with the lower particle density,the recognition rate of our method can reach 92.6%,in the area with the higher particle density,the recognition rate can alsoreach 86%,it is higher than the recognition rate of method combined gradient and gray respectively by 9.9% and 42.7%.The proposed algorithm is also more robust to the light field and the imaging effect.

关 键 词：异向导电胶膜 粒子重叠 K均值 图像熵 凸性 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]