交通运输系统工程与信息

交通运输系统工程与信息 ›› 2021, Vol. 21 ›› Issue (2): 37-43.

• 智能交通系统与信息技术 • 上一篇    下一篇

基于自动车牌识别数据的混合交通流饱和流率实时估计

王殿海，郭佳林，蔡正义*   

  1. 浙江大学，建筑工程学院，智能交通研究所，杭州 310058
  • 收稿日期:2020-11-07 修回日期:2021-01-15 出版日期:2021-04-25 发布日期:2021-04-25
  • 作者简介:王殿海(1962- )，男，吉林大安人，教授，博士。
  • 基金资助:

    国家自然科学基金/National Natural Science Foundation of China(61773338，71901193，52072340)。

Real Time Estimation of Saturated Flow Rate of Mixed Traffic Flow Based on Automatic License Plate Recognition Data

WANG Dian-hai, GUO Jia-lin, CAI Zheng-yi *   

  1. IITS, College of Civil Engineering and Architecture, Zhejiang University, Hangzhou 310058, China
  • Received:2020-11-07 Revised:2021-01-15 Online:2021-04-25 Published:2021-04-25

PDF

摘要：

为解决混合交通流饱和流率测算的实时性和时变性问题，实时获得混合交通流的饱和流率用以信号配时，本文提出基于自动车牌识别数据(Automatic License Plate Recognition，ALPR)的混合交通流饱和流率实时自动估计方法。首先，分信号周期提取车头时距数据，在当前车和后车车辆类型确定时车头时距满足同一正态分布的假设基础上，构建车头时距的高斯混合模型并应用 EM(Expectation Maximization) 算 法 求 解 ；其 次 ，基 于 赤 池 信 息 准 则 (Akaike Information Criterion，AIC)选取高斯混合模型的最优个数，拟合数据得到高斯混合模型参数；最后，根据车头时距的高斯混合模型推算出混合交通流饱和流率。以杭州城市道路3条路段的ALPR数据为例，分析基于 ALPR 数据获取车头时距的采样误差，对模型进行验证，并与传统的 HCM(Highway Capacity Manual)方法进行对比。结果表明：基于ALPR数据的车头时距采样误差满足精度要求； 与HCM的实测法相比，模型所得的混合饱和交通流率相对误差小，结果准确；该方法与传统的标准车流饱和流率折算法效果相近，并考虑混合交通流时变特性，能自动部署实时计算，鲁棒性良好，有实际应用意义。

关键词: 智能交通, 饱和流率, 高斯混合模型, 车牌识别数据, 交通控制

Abstract:

In order to solve the problem of real-time and time- varying of saturated flow rate measurement of mixed traffic flow, a real-time saturated flow rate estimation method based on automatic license plate recognition (ALPR) is proposed in this paper. Firstly, the headway data is extracted in different signal periods. With the assumption that the headway follows the identical normal distribution when the types of successive vehicles are determined, the Gaussian mixture model of vehicle headway is constructed and solved by EM algorithm. The parameters in the Gaussian mixture model are obtained by fitting the data with the best clustering number based on Akaike information criterion. According to the Gaussian mixture model of headway, the saturated flow rate of mixed traffic flow can be calculated. Taking the ALPR data in three sections of Hangzhou urban road as an example, the sampling error of headway obtained from ALPR data is analyzed and compared with the traditional HCM (highway capacity manual)method. The results show that the relative error of mixed saturated traffic flow rate obtained by the model is small. The sampling error of headway based on ALPR data meets the accuracy requirements. Compared with the traditional HCM method, the method can automatically deploy real- time calculation considering the time- varying characteristics of mixed traffic flow, which has practical significance and good robustness.

Key words: intelligent transportation, saturated flow rate, Gaussian mixture model, license plate recognition data, traffic control

中图分类号: