一种基于深度学习的离散化交通状态判别方法

发布时间：2018-01-31 22:54

本文关键词： 智能交通交通状态判别深度学习交通状态离散化交通状态编码　出处：《交通运输系统工程与信息》2017年05期 　论文类型：期刊论文

【摘要】：在智能交通信号控制和交通流诱导系统中,交通环境状态的有效判别是影响交通控制决策的先决条件,本文针对交通流产生的大数据信息,结合深度学习算法提出一种离散化交通状态的判别方法.给出了包括交通状态数据采集、状态数据描述、状态深度学习和判别等功能模块的系统架构,构建了一种离散交通状态编码方法,为深度学习交通状态特征提供了数据基础.模型训练阶段,对采集到的二值和连续值交通状态数据,分别构建了两种不同的深度置信网络实现交通状态特征的无监督学习;模型微调阶段,在整合形成的高层抽象特征向量顶端增加softmax分类器,采用反向传播算法实现参数微调.最后,该方法基于VISSIM微观交通软件进行仿真,实验结果表明,离散交通状态编码方法可有效表达交通状态,基于深度学习的交通状态判别方法相对传统方法具有较高的准确度.
[Abstract]:In the intelligent traffic signal control and traffic flow guidance system, the effective identification of traffic environment is a prerequisite for traffic control decision-making. This paper aims at the big data information generated by traffic flow. Combined with depth learning algorithm, a discrete traffic state identification method is proposed, and the system architecture including traffic state data collection, state data description, state depth learning and discrimination is presented. In this paper, a discrete traffic state coding method is constructed, which provides a data base for the in-depth learning of traffic state characteristics. In the training stage of the model, the collected binary and continuous traffic state data are obtained. Two different depth confidence networks are constructed to realize unsupervised learning of traffic state characteristics. At the stage of model fine-tuning, softmax classifier is added at the top of the integrated high-level abstract feature vector, and parameter fine-tuning is realized by back-propagation algorithm. Finally. The simulation results based on VISSIM microcosmic traffic software show that the discrete traffic state coding method can express the traffic state effectively. Compared with the traditional method, the traffic state discrimination method based on depth learning has higher accuracy.
【作者单位】：中国海洋大学信息科学与工程学院;山东理工大学计算机科学与技术学院;
【基金】：淄博市校城融合发展计划(2016ZBXC003) 国家重点研发计划(2016YFB1001103)~~
【分类号】：U491
【正文快照】： 0引言随着社会城市化进程加快,全球大中城市均面临交通拥堵压力.同样,我国家用汽车保有量近年来持续快速增长,城市交通拥堵已成为大众出行讨论焦点,我国政府也采取多方措施疏堵、治堵.为解决交通拥堵问题,通过信息技术手段判别交通环境状态,适时调整交通信号控制策略提高通行

【相似文献】