基于 Hadoop 的海量图像检索系统

摘要

In a traditional single-node architecture image retrieval system,facing the problems of slow retrieval speed,poor concurrency etc when retrieved massive image data,proposed an image retrieval method based on the Hadoop,combining image retrieval technology with MapReduce frame,image feature database stored in the distributed file system HDFS,computing nodes using the scheduling algo-rithm based on Hadoop distributed storage,enhanced concurrent processing capability for multiple data,at the same time compressed the calculated data. Test and experiment results show that,the method in dealing with large data retrieval,compared with the single-node re-trieval system,can effectively reduce the search time,improve the retrieval speed.%　　在传统图像检索系统中，由于采用单节点架构，面对海量图像数据检索时存在检索速度慢、并发性差等问题。文中提出了一种基于 Hadoop 的图像检索方法，将图像检索技术与 MapReduce 框架相结合，图像特征库存储于分布式文件系统 HDFS 中，计算节点采用基于 Hadoop 的分布式存储调度算法，增强对多数据的并发处理能力，同时对计算后的数据进行压缩处理。实验表明，该方法在处理大数据图像检索时，与单节点检索系统相比，能够有效降低检索时间，提高检索速度。