基于图像特征索引的并行检索技术研究  

Parallel Retrieval System Based on Image Feature Index

作　　者：蔡晓东[1] 华娜[1] 吴迪[1] 陈文竹[1] 

机构地区：[1]桂林电子科技大学信息与通信工程学院,广西桂林541004

出　　处：《电视技术》2015年第13期24-26,113,共4页Video Engineering

基　　金：国家"863"计划项目(2012BAH20B01;2014BAK11B02);广西自然科学基金项目(2012GXNSFAA053232;2013GXNSFAA019326);广西高校科学技术研究项目(2013YB092)

摘　　要：为了提高建立索引、检索图像的速度,提出云架构上基于图像特征索引的并行检索系统。该检索系统主要有3个模块:海量小图片分布式存储(Store)、并行建立图像特征索引(Indexing)、并行图像检索(Retrieve)。在Store模块中提出针对海量图片的合并存储,Indexing模块中提出索引缓存模式,避免重写索引的输出接口,Retrieve模块中对索引进行分片管理,以及并行检索。实验结果表明,相对于其他图像检索系统,基于图像特征索引的检索系统有效减少了图像特征索引建立时间,缩短了图像的检索时间,提高了图像检索速率。In order to improve the speed of indexing and retrieving images, the parallel retrieval system based on image features index in cloud is proposed. The retrieval system mainly consists of three parts, the distributed store of massive small images（ Store）, building index of image features in parallel（Indexing） , image retrieval in parallel （Retrieve）. In the part of store, the combined storage for massive small images is proposed, in the part of Indexing, in order to avoid rewriting the output interface of index, the mode for caching index is proposed, in the part of Retrieve, by the fragmented management for features index, it proposes the image retrieval in parallel. The experimental results show that the retrieval system reduces the time for indexing and retrieving effectively, it improves the speed for image retrieval.

关 键 词：分布式存储 特征索引 并行检索 分片管理 

分 类 号：TN911.73[电子电信—通信与信息系统] TP3[电子电信—信息与通信工程]