基于Hadoop平台的图像检索分布式算法的改进研究  

Research and Improvement of Distributed Algorithm for Image Retrieval Based on Hadoop Platform

作　　者：丁灿 侯春萍[1] 王宝亮[1] 

机构地区：[1]天津大学电子信息工程学院,天津300072

出　　处：《南开大学学报（自然科学版）》2017年第4期46-51,共6页Acta Scientiarum Naturalium Universitatis Nankaiensis

摘　　要：针对单机环境下图像检索算法存在的计算瓶颈问题,提出1种基于Hadoop平台Map-Reduce框架的图像检索改进算法.首先利用云存储和分布式算法的优点,对图像检索算法进行并行化设计,然后通过改进的K-Means和Bo F算法,对原有图像检索算法进行优化,提高检索效率和准确性.实验结果表明,提出的方法能够很好地提高图像检索算法的执行速度,同时获得较好的图像检索准确性.A new image retrieval algorithm based on Hadoop platform Map-Reduce framework is proposed to calculate the bottleneck of image retrieval in a single machine environment.Firstly,the advantages of cloud storage and distributed algorithm are used to design the image retrieval algorithm.Then,the K-Means and Bo F are improved,and the original image retrieval algorithm is optimized to improve the efficiency and accuracy.Experimental results show that the method proposed can improve the performance of image retrieval algorithm,and obtain a better accuracy of image retrieval.

关 键 词：分布式运算 图像检索 MAP-REDUCE 尺度不变特征变换 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]