AMOLED不良MAP自动缺陷定位方法研究

Research on auto MAP defect location algorithm of AMOLED

AMOLED生产工艺复杂,在生产过程中容易受环境、化学气体、液体洁净度及设备工艺参数的影响,在玻璃面板上形成大小不均的点状不良。当面板上出现大量不良点聚集时,会导致该面板直接报废,影响产品的最终良率。本文方法结合机器学习与图像处理技术,可实现不良坐标点MAP图的在线智能分析。采用层次聚类算法对不良坐标点进行聚集分类,对每类离散点簇采用Alpha Shapes算法提取其外轮廓点;通过图像区域拟合算法,拟合出每类点簇的最小包围图形区域,并计算图形的Hu几何矩,推导出各区域的区域特征(区域点密度、面积、区域的质心、区域方向、区域长宽比等),并利用区域特征对图形区域进行筛选,最终定位出不良点的目标聚集区域。实验证明,本文方法可实现MAP图智能在线分析,自动定位不良点聚集区,可替代当前人工目检的方式,保证检测质量,降低工厂成本。

Due to the complicated production process,AMOLED is easily influenced by the cleanliness of environment,chemical gas and liquid which will cause a large amount of uneven dot defects on glass during manufacturing.Concentrated region,that gathers a large amount of dot defects,will lead to final yield loss.As a result,detecting and locating the concentrated region timely is very important in AMOLED manufacturing field.Firstly,hierarchical clustering algorithm is used to divide the defects into classes with Euclidean distance threshold of defect points and alpha shapes algorithm is adopted to extract the outer contour points.Secondly,the smallest surrounding region of each class is fitted by interpolation algorithm,and then the region features is calculated by Hu moment algorithm,such as region density,centroid,orientation,area,length-width ration and etc.Finally,the candidate regions are filtered according to the feature value of destination regions.Experimental result shows that the proposed algorithm can realize intelligent analysis of defect MAP image,and locate the defect concentrated regions automatically which can replace manual inspection to guarantee quality and lower cost.