基于Hadoop平台图像识别.docVIP

基于Hadoop平台图像识别 　　摘 要： 基于Hadoop平台以字符识别为例建立图像识别系统。所设计的系统在借鉴云平台高扩展性以及高效性等优势的基础上，有效地解决了传统字符识别系统在计算效率以及数据处理方面所存在的不足。通过实例验证了基于Hadoop平台进行图像识别相比单机图像识别系统具有更高的效率：在仅具有2个节点的Hadoop图像识别平台上进行字符图像的识别时，由于节点数较少，在2台计算机中消耗的数据交换时间使得Hadoop图像识别平台进行图像识别的总时间甚至超过了单台计算机所使用的时间，而在具有4个节点、6个节点和8个节点的Hadoop图像识别平台上，处理相同图像所使用的时间随着节点数量增多而降低 关键词： 字符识别； Hadoop平台； 图像识别； 数据交换时间 中图分类号： TN911?34； U495 文献标识码： A 文章编号： 1004?373X（2017）04?0128?04 Image recognition based on Hadoop platform ZHAO Zhen （Department of Software Engineering， Inner Mongolia Electronic Information Vocational Technical College， Hohhot 010000， China） Abstract： The image recognition system was established based on Hadoop platform， which takes the character recognition as an example. The system based on the advantages of good scalability and high efficiency of the cloud platform can effectively eliminate the shortcomings of the traditional character recognition system in the aspects of computing efficiency and data processing. The fact that the efficiency of the image recognition system based on Hadoop platform is higher than that of the stand?alone image recognition system is verified with an instance. The data exchange time consumed in two computers makes the total time of the image recognition based on Hadoop image recognition platform with only two nodes longer than the use time of the image recognition based on single computer due to the less node quantity， when the character image is recognized on Hadoop image recognition platform with two nodes. The use time for processing the same image on Hadoop image recognition platform with four nodes， six nodes or eight nodes is deduced with the increase of the node quantity. Keywords： character recognition； Hadoop platform； image recognition； data exchange time 在??互联网图片进行匹配和分类的过程中，单台计算机已经无法满足相应的需求，而分布式计算框架能够稳定和高效地匹配和分配大量的互联网图片，具有显著的优势[1?2]。本文基于Hadoop平台，以字符识别为例建立图像识别系统。本文所设计的系统在借鉴云平台高扩展性以及高效性等优势的基础上有效的解决在计算效率以及数据处理方面传统字符识别系统所存在的不足。主要体现在以下两个方面：利用Hadoop能够在平台各个节点中分配字符识别任务，对于任