【作者】 于万霞；

【导师】 杜太行；

【作者基本信息】 河北工业大学 ， 电机与电器， 2008， 博士

【摘要】 现代城市交通的智能控制是ITS的重要组成部分,而交叉口是决定道路通行顺畅的关键。单交叉口的实时控制是交通控制系统的基础。城市交通的智能控制实现的前提和关键是实时准确的交通流量预测。全面、准确的采集交通信息是实现交通智能化的基本保障,交通流量预测的准确性也取决于数据样本的准确性。因此,发展城市交通的智能化技术,研究城市单交叉口交通的智能控制、流量预测以及交通信息采集技术成为今后ITS的发展方向。本文对城市单交叉口交通信号的智能控制、流量预测和基于视频的图像采集技术进行了研究和探讨。首先,提出了城市智能交通的整体结构设计,即包括车流量采集、流量预测和交通信号控制几个模块;然后,对每个模块的具体设计进行了详细的介绍。车流量采集采用基于视频的图像处理技术,采用建立在YCbCr色彩空间上的背景帧差法进行图像的分割处理,采用数学形态学、图像的连通性等进行图像的去噪,并提出了计算梯度结合峰值的方法对车辆进行计数。车流量预测模块中,在分析车流量预测中存在的问题和交通流特性的基础上,建立了模糊神经网络预测模型,并提出了采用蚁群和粒子群结合的方法优化模糊神经网络参数。算法中,将蚁群和粒子群组成主从结构,其中,蚁群在全局解空间进行搜索,粒子群在局部解空间进行搜索,并将解反馈给主级。交通信号控制模块中,以多相位单交叉口的信号灯为控制对象,建立了模糊神经网络的交通信号控制模型,并采用粒子群优化模糊神经网络参数。最后,对基于DSP的交通信号控制器进行了软硬件设计。仿真结果表明,本文提出的算法有效地提高了车流量检测和流量预测的精度及控制的效率。更多还原

【Abstract】 In modern times, the intelligent control of urban traffic is an important part of ITS. The intersection acts as the key factor in deciding the road traffic. The traffic control system bases on real time control to isolated intersection. The precondition and key of intelligent control of urban traffic is real time and exact traffic flow prediction. Collecting the exact and complete traffic information is the basic ensure of finishing intelligent traffic. Besides, the veracity of data samples decides the veracity of predicting traffic flow. Therefore, developing the intelligent technology of urban traffic, researching the following technologies: intelligent traffic control of urban isolated intersection, traffic flow prediction and traffic parameters collection are the development direction of ITS.The intelligent traffic control of urban iaolated intersection, flow prediction and image collection technologies are researched and discussed in this paper. Firstly, the paper puts forward the whole structure design of urban traffic, which includes three modules: collecting vehicle parameters, predicting vehicle flow and controlling traffic signal. Secondly, the detailed design of each module is introduced. In the module of collecting vehicle parameters, the image processing technology based on video is adopted. The paper adopts background difference based on YCbCr color space to divide up the image, and math morphologic and the connectedness of image to eliminate the noise of image. The way of grads extremum is putted forward to check the vehicle flow. In the module of predicting vehicle flow, after analyzing the existent question in traffic flow prediction and the characteristic of traffic flow, a fuzzy neural network(FNN) prediction model and the learning algorithm of FNN based on the associative way of ant colony optimization(ACO) algorithm and particle swarm optimization(PSO) algorithm.are putted forward The learning algorithm is formulated in a form of hierarchical structure. The global search is performed by ant population at the master level, while the local search is carried out by particle population at the slave level and the best solution is fed to the ant population. In the module of controlling traffic signal, the paper establishes a traffic signal control model adopting FNN and the learning algorithm of FNN based on PSO. Finally, the traffic signal controller based DSP is designed, the hardware and software designs are presented. The simulation results demonstrate the proposed models can improve accuracy in taking count of vehicle flow, predicting vehicle flow and controlling traffic signal.更多还原