基于MapReduce的液晶屏缺陷检测方法.pdf

202 2017 ，53（5） Computer Engineering and Applications 计算机工程与应用 基于MapReduce 的液晶屏缺陷检测方法 夏晓云，张仁斌，谢 瑞，王 聪 XIA Xiaoyun, ZHANG Renbin, XIE Rui, WANG Cong 合肥工业大学 计算机与信息学院，合肥 230009 School of Computer and Information, Hefei University of Technology, Hefei 230009, China XIA Xiaoyun, ZHANG Renbin, XIE Rui, et al. MapReduce approach for defect inspection of TFT-LCD. Computer Engineering and Applications, 2017, 53 （5 ）：202-206. Abstract: Various types of defects would come into being in the process of producing 6th TFT-LCD. There are bottlenecks of storing resources and calculating time by inspecting defects with one single machine. It is a new way to deal with mas- sive high-resolution LCD images by Hadoop clusters, which has the advantage in computing and storage capacity. Since the defect is a local feature of high-resolution LCD image, a distributed method of defect inspection based on MapReduce framework is proposed in this paper. To solve the problem of low efficiency of high-resolution image in defect inspection, the approach can be simply described as follows. First, the high-resolution image is split into multiple small splits, which are parallel inspected in the following step. In the final step, the intermediate results are aggregated to obtain the final result. The experimental results show that this approach can inspect defects simultaneously on Hadoop cluster with a good speed up rate. Key words: Hadoop; high-resolution image; defect inspection; MapReduce 摘 要：5th 液晶屏在生产过程中会产生多种类型的缺陷，通过单一节点进行缺陷检测存在存储资源和计算时间的瓶 颈。利用Hadoop 集群的分布式计算、存储能力处理海量的高分辨率液晶屏图像是一个新的思路。针对高分辨液晶 屏图像缺陷局部性特点，设计基于MapReduce 的分布式缺陷检测方法，对高分辨率图像分块，并行完成每块图像的 缺陷检测，再将检测结果归并，从而解决高分辨率图像缺陷检测效率低下问题。通过运行在Hadoop 平台上的实验 表明，该方法在完成缺陷检测的同时具有良好的效率提升。 关键词：Hadoop ；高分辨率图像；缺陷检测；MapReduce 文献标志码：A 中图分类号：TP317.4 doi ：10.3778/j.issn. 1002-8331.1507-0286 1 引言 存储资源瓶颈，以及一些其他的问题都影响并行计算的 液晶屏在生产过程中由于技术原因，会产生多种类