交通流大数据中的套牌车并行检测算法  

A Parallel Algorithm for Detecting Fake Plate in Big Data of Traffic Flow

作　　者：王涛[1] 王顺[1] 沈益民[2] 

机构地区：[1]滁州学院计算机与信息工程学院,安徽滁州239000 [2]成都工业学院计算机系,四川成都611730

出　　处：《湖北工程学院学报》2014年第6期29-32,共4页Journal of Hubei Engineering University

基　　金：安徽省科技攻关计划项目(1401b042013);四川省科技厅应用基础计划项目(2013JY0059);滁州学院科研启动项目(2012QD07)

摘　　要：传统的套牌车识别算法通过串行工作方式在网格化城市交通监控系统所产生的大规模数据中进行两两比对实现套牌车检测,因此在处理海量数据时存在性能瓶颈问题。提出了一种新的基于Hadoop的MapReduce算法模型,该算法具有并行特征,通过引入多台硬件计算资源协同处理大规模数据下的套牌车检测问题,显著提高了计算性能。同时,采用基于动态旅行时间实时的时空窗口计算技术,能进一步提高算法的检测速度和识别精度。Traditional fake plate detecting algorithms, which typically utilized pair-wise comparisons todetect fake plate in traffic flow big data produced by urban grid monitoring systems, were performedin a serial manner. Therefore, the detecting systems were bottle-necked when processing massive traf-fic data. In this paper,a new Hadoop-MapReduce based method was proposed, which could be seen asa parallel algorithm. With more computers, the newly proposed method could improve the perform-ance of the detecting program for fake plate detection. Moreover,a dynamic-travel-time based spatial-temporal window was used to promote the detecting speed as well as accuracy.

关 键 词：交通流大数据 套牌车 MAP REDUCTION HADOOP 

分 类 号：TN927[电子电信—通信与信息系统]