摘要
第六代液晶屏在生产过程中会产生多种类型的缺陷,通过单机进行缺陷检测存在存储资源和计算时间的瓶颈。利用Hadoop集群优势处理海量的高分辨率液晶屏图像是一个新的思路。针对线阵CCD(charge-coupled device)相机采集特点,提出一种基于MapReduce的分布式缺陷检测方法,使用改进的C-V图像分割模型,完成液晶屏模糊边缘的缺陷分割,对处于子图边缘的缺陷进行二次缺陷提取,提高缺陷检测的准确率,并采用SVM(support vector machine)分类器完成缺陷的分类。实验表明,该方法提高检测效率的同时降低了缺陷的误判率,为分布式存储分块图像、缺陷测量等相关研究奠定了基础。
Various types of defects would come into being in the process of producing 6th TFT-LCD. Inspecting defects with one single machine would face the bottlenecks of storing resources and calculating time. It is a new way to deal with massive high-resolution LCD images by Hadoop clusters, which has the advantages in computing and storage capacity. In consideration of the characteristic of collecting images by CCD （ charge-coupled device） camera, this paper proposed a MapReduce approach for distributed defect inspection. First, it used the improved C-V model to segment the defects of TFT-LCD with blurry con- tour. Then, it executed the second segmentation for the defect at the edge of the sub image to improve the accuracy of defect inspection. Finally, it used the SVM classifier to classify defects. Experiment demonstrates that this method can inspect de- fects simultaneously with a good speed up rate, and reduce the miscarriage of justice for defects. In addition, it lays the foun- dation for the research on the distributed storage of image splits and defect measurement and so on.
出处
《计算机应用研究》
CSCD
北大核心
2016年第8期2534-2538,2542,共6页
Application Research of Computers
基金
国家自然科学基金资助项目(61271121)
