基于对数变换的交通流的研究

发布时间：2018-07-02 19:24

本文选题：交通流预测 + 低阶非线性变换　；参考：《天津大学》2014年硕士论文

【摘要】：随着社会的进步和经济的发展,人们对于交通的需求日益增长。但是,现有交通道路的状况与交通流量的增长之间的矛盾与日俱增。这就需要一个可以对交通流量进行实时、准确的预测,用来实现对交通的控制和诱导。经过理论研究和实践表明,智能交通系统可以较好的解决这一问题,其主要的方法是基于交通流参数预测和交通事件检测。所以,找到交通流中较好的参数和交通事件中较好的检测方法是亟待解决的问题。本论文的主要研究内容和成果如下:(1)将一种低阶非线性变换(对数变换)运用到小波阈值去噪过程中。在研究中,通过对标准信号加上异常点或者跳跃点,不采用对数变换时,抑制异常值的范围是[-0.7340,0.4314];经过对数变换后,使得抑制异常点的范围变化为[-0.7428,0.4685]。区间长度从1.1654增加到1.2113,增加了3.94%,说明通过非线性变换后,对小波阈值去噪有一定效果,并针对这一现象进行了进一步的理论分析。该研究属于“机理+辨识”预测策略中的非平稳数据的平稳化方法研究。(2)给出了对数变换提高小波阈值去噪效果的数学分析,并通过对数变换下的泰勒展开式反映了对数变换对于小波阈值去噪的影响,即:对数变换降低非平稳时间序列预测的均方根误差(MSE,Mean Squared Error),但引起平均误差(ME,Mean Error)轻微的增加。并且使用数值试验和公路交通流预测的实例对其进行了验证与分析。(3)通过对交通流信号进行去噪的新旧方法对比,和对五种预测模型直接进行预测和经过非线性变换之后再进行预测两种方法的对比,说明非线性变换对于小波去噪是有良好效果的。经过非线性变换之后再进行去噪,使得信号得到了较大的改善。
[Abstract]:With the progress of society and the development of economy, people's demand for transportation is increasing day by day. However, the contradiction between the existing traffic conditions and the growth of traffic flow is increasing day by day. This requires a real-time and accurate prediction of traffic flow, which can be used to control and guide traffic. The theoretical research and practice show that the intelligent transportation system can solve this problem well. The main methods are based on traffic flow parameter prediction and traffic event detection. Therefore, it is an urgent problem to find better parameters in traffic flow and better detection methods in traffic events. The main contents and achievements of this thesis are as follows: (1) A low order nonlinear transform (logarithmic transform) is applied to the wavelet threshold denoising process. In the study, the range of suppression outliers is [-0.7340 ~ 0.4314] by adding outliers or jumping points to the standard signals, and the range of the suppressed outliers is [-0.7428 (0.4685)] after logarithmic transformation. The interval length is increased from 1.1654 to 1.2113, and 3.94 is added. It shows that the wavelet threshold de-noising is effective after nonlinear transformation, and the further theoretical analysis is carried out in view of this phenomenon. This research belongs to the stationary method of non-stationary data in the prediction strategy of "mechanism identification". (2) the mathematical analysis of logarithmic transformation to improve the denoising effect of wavelet threshold is given. The Taylor expansion under logarithmic transformation reflects the effect of logarithmic transformation on wavelet threshold denoising, that is, the logarithmic transformation reduces the MSEM mean square error of the prediction of non-stationary time series, but causes a slight increase in the mean error. Numerical experiments and highway traffic flow prediction examples are used to verify and analyze it. (3) by comparing the new and old methods of traffic flow signal denoising, Compared with the two methods of forecasting five prediction models directly and after nonlinear transformation, it is shown that nonlinear transform has good effect on wavelet denoising. After nonlinear transformation, the signal is de-noised, and the signal is improved greatly.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.112

【参考文献】