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。
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 im?age 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 rec?ognition based on single computer due to the less node quantity,when the character image is recognized on Hadoop image recog?nition 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.
作者
赵祯
ZHAO Zhen(Department of Software Engineering,Inner Mongolia Electronic Information Vocational Technical College,Hohhot 010000,China)
出处
《现代电子技术》
北大核心
2017年第4期128-131,共4页
Modern Electronics Technique
基金
国家自然科学基金(41261050)
