高速公路交通事件自动检测算法  

Automatic incident detection algorithm for expressways

作　　者：李琦 姜桂艳[2] 

机构地区：[1]青岛市城市规划设计研究院,山东青岛266071 [2]宁波大学海运学院,浙江宁波315211

出　　处：《哈尔滨工程大学学报》2013年第9期1193-1198,1208,共7页Journal of Harbin Engineering University

基　　金：国家自然科学基金资助项目(51278257);高等学校博士学科点专项科研基金资助项目(20110061110034);浙江省自然科学基金资助项目(LY12F01013)

摘　　要：为解决目前我国高速公路交通检测器布设数量严重不足所导致的交通事件检测效果不佳的问题,在分析了收费数据特征的基础上,设计了一种基于收费数据的交通事件自动检测算法.该算法以标准偏差法为基础,首先为了减少因交通波动引发的误警,提出了一种基于滚动时间序列的交通参数合成方法;在此基础上,为了减少因常发性交通拥挤引发的误警,提出了一种综合考虑交通参数数据横向时间序列和交通参数数据纵向时间序列的改进方案;进而,为了减少因算法自身的检测逻辑引发的误警,提出了一种基于数据分析时间窗口内的交通参数标准差以及当前采样间隔交通参数相对于其以前平均值改变程度的改进方案.采用我国浙江省沪杭甬高速公路的实测收费数据进行验证和对比分析的结果表明,在相同的误警水平下,本文算法的检测率明显优于标准偏差法,平均检测时间与标准偏差法基本持平,且本文算法具有良好的鲁棒性.In order to solve the problem of ineffective incident detection due to the severe shortage of traffic sensors for expressways in China, on the basis of analyzing toll data characteristics, an automatic incident detection algo-rithm using toll collection data was designed. The algorithm was based on standard normal deviation algorithm. First, in order to reduce the false alarms caused by traffic fluctuations, this paper proposed a traffic data synthetic method based on rolling time series. On the basis of the first step, in order to reduce the false alarms caused by re-curring congestion, this paper proposed a modification by comprehensively considering the horizontal time series and the longitudinal time series of traffic parameter data. Furthermore, in order to reduce the false alarms caused by detection logic of the algorithm itself, this paper proposed an improved scheme based on the standard deviation val- ue of traffic parameters and the current traffic flow minus the mean in the data analyzing time window. The proposed algorithm was tested with field data collected from the Hu-Hang-Yong Expressway in China. The test and compari- son analysis results indicate that at the same false alarm rate level, the detection rate of the proposed algorithm is significantly better than the standard normal deviation algorithm, the mean time of detection is basically equivalent to that of the standard normal deviation algorithm. Moreover, the proposed algorithm has very strong robustness.

关 键 词：交通运输系统工程 交通事件自动检测 收费数据 标准偏差法 

分 类 号：U491[交通运输工程—交通运输规划与管理]