海量车牌识别数据集上基于时空划分的旅行时间计算方法

城市路段旅行时间计算是智能交通领域的一个研究热点.车牌识别数据作为近年来新兴的一种针对城市道路行驶车辆的实时监测数据,具有持续生成且数据量大、时间空间相关等特性.为了利用车牌识别数据集进行高效、准确的旅行时间计算,给出了基于车牌识别数据集的旅行时间计算定义,在此基础上提出一种基于时空划分的流水线式并行计算模型,并给出了该模型基于实时MapReduce的实现.通过一组基于海量真实车牌识别数据集的实验表明,本文方法在亿级车牌识别数据集上的旅行时间计算性能方面相对于直接基于Hadoop的实现可以提高3倍以上,同时具有适合细粒度划分及受路网规模影响小的特点.

A Travel Time Calculation Method Based on Spatio-Temporal Data Partition of Large-Scale License Plate Recognition Data Set

The calculation of travel time of city roads is an important issue in the domain of the intelligent transporta-tion system research.License plate recognition data is one kind of monitoring data for vehicles running on urban roads,which has some new features,such as high volume,high velocity and spatio-temporal correlation.In order to achieve travel time calculations on massive license plate recognition data collection,we present the formal definition of travel time calculation based on license plate recognition data set,and propose a pipelined parallel computing model based on spatio-temporal data partition.Moreover,the implementation of the computing model is given based on a real-time MapReduce computing sys-tem.The corresponding experiments based on real license plate recognition data set show that,the computing performance on million-level data sets of our method can achieve three times increasing compared to traditional travel time calculation meth-ods.Meanwhile our method is more suitable for fine-grained partition and large scale traffic network.