自动识别环境下车辆的出行矩阵估计新方法

鉴于以视频牌照识别系统为代表的车辆自动识别(automatic vehicle identification,AVI)技术在我国逐步应用的现实,提出了利用AVI检测信息估计高精度车辆起讫点矩阵(OD- matrix)的新方法.该方法首先将检测的车辆信息分为4类(起讫点已知、起点或终点及部分路径已知、仅知起点或终点、仅知部分路径),然后利用第1类信息根据AVI检测误差直接扩样更新基础OD矩阵；利用第2,3,4类信息,参照粒子滤波算法思想,基于贝叶斯估计理论修正更新路段-路径流量关系,进而用蒙特卡罗随机过程确定可能路径以及OD;最后根据AVI获得的路径流量信息反向验算校正OD.根据上海市目前视频牌照识别系统的应用现状,选择以南北高架快速路为研究对象,根据牌照识别系统检测的动态车辆信息,对布设9个视频检测器的南北高架沿线17个出入口的OD进行了估计应用.结果表明,在路网仿真模型误差≤15％、AVI设施覆盖率为27.2％以及检测误差在10％的前提下,运用本方法,OD估计的总体平均相对误差仅为11.09％.该方法能充分利用AVI检测的个体车辆不完整路径信息,且计算效率高,可满足实际动态交通管理的需求.

A New Method of OD Estimation Based On Automatic Vehicle Identification Data

With the development and application of video license plate recognition system which represented the automatic vehicle identification(AVI) technologies in China,a novel high resolution OD estimation method was proposed based on AVI detector information.4 detected categories(Ox+Dy,Ox/Dy+(s),Ox/Dy、P(s)) were divided at the first step.Then the initial OD matrix was updated by using the Ox+Dy sample information considering the AVI detector errors.Referenced by particle filter,the link-path relationship data were revised by using the last 3 categories information based on Bayesian inference and the possible trajectory and OD were determined with the Monte Carlo random process.Then the OD was corrected through reserve operation based on AVI path flow information at last.Finally,according to the current application of video detector in Shanghai,the North-South expressway was selected as the test bed which including 17 OD pairs and 9 AVI detectors.The results show that the calculated average relative error is 11.09% under the constraints that the simulation error is under 15%,the detector error is about 10% and the AVI facility coverage is 27.2%.This method is highly efficient and can make full use of the partial vehicle trajectory which can be satisfied with the dynamic traffic management application in reality.