Font Size: a A A

Traffic Incidents Detection Based On Traffic Low

Posted on:2015-09-22Degree:MasterType:Thesis
Country:ChinaCandidate:Y LiFull Text:PDF
GTID:2272330452458948Subject:Control Engineering
Abstract/Summary:PDF Full Text Request
With the growth of automobile amount in china, traffic safety, traffic congestionand environmental pollution became a major problem in traffic field. Intelligenttransportatiom system has been widely developed as the important method to fix theseissue. Traffic incident detection is the core of intelligent transportatiom system andhas deeply impact to the development of the intelligent transportatiom system. Thereal-time, accurate and fast method of traffic incident detection is greatly significant.The main contents and results of this paper are as follows:(1) Because the the low-order non-linear transformation can improve thesignal-to-noise ratio of the input signal, reduce the influence of the "bad" datum in thecomplex time series and improve the decomposition between noise and signal, thistransformation is applied into the forecast field of traffic flow includingautoregressive model, moving average model, exponential smoothing model andcombination forecasting model combined by simple average method and the variancereciprocal weighting method. The empirical results show that the low-order non-lineartransformation can improve the accuracy of the individual forcast and the combinationforecast.(2) The low-order logarithmic transformation is applied to wavelet denoising area.The experiments compare the effect of wavelet denoising with the low-orderlogarithmic transformation and the effect of wavelet denoising without the low-orderlogarithmic transformation. The results show that the low-order logarithmictransformation can reduce the sensibility of wavelet denoising.(3) The wavelet denoising method and two kinds of fourier denoising are appliedto the traffic incident. The traffic flow is decomposed to signal series plus noise series.The distribution of the detected time point of the traffic flow and the confidence levelof the occurrence of traffic incident can be caculated by the analysis of the signal andnoise series. And then, the probability of traffic congestion incident can be acquiredby the confidence level.
Keywords/Search Tags:Traffic incident detection, Low-order non-linear transformation, Combination forecast, “Mechanism+identification” forecasting strategy
PDF Full Text Request
Related items