Traffic State Identification Based On Multi Source Information Processing Technique

With the rapid growth of urban car ownership,city's available spatial resources are decreasing.Therefore the imbalance between the supply and demand of road traffic development has become increasingly prominent,and leading to the gradually serious problem of traffic congestion.The rapid and accurate identification of traffic state for the urban road network is a prerequisite for the development of mitigation strategies.The traffic information quality directly affects the accuracy of the resultant traffic state,due to the single detector data failing to meet the requirements of reliability calculations.It is therefore essential that,multi-source data is processed to improve the quality of data.In this paper,traffic data processing and traffic state identification are studied:(1)Targeting the problem that the detector is missing data which affects the quality of the entire data set.This paper proposes a combined data inpainting model based on the Grey Theory and Support Vector Machine(SVM).This model takes into account the advantages of the Grey Theory in dealing with the data issue of small samples and poor information,repairing traffic flow data faults while requiring fewer data.At the same time,taking into account the SVM for processing nonlinear data provides advantages and good generalization ability,it will therefore be used to repair the missing data traffic flow,and prove that the combined model repair precision method offers superior performance.(2)In consideration of the advantages and disadvantages of the traditional traffic detector in the detection of traffic parameters,a multi-source traffic data fusion model based on Genetic Wavelet Neural Network is proposed.The model is based on different detector data sequences,first,from a single data source to construct the genetic wavelet neural network traffic prediction model.The results of which are fused by the least squares fusion method,consequently the quality of data is improved than that of the single detector verification after data fusion.(3)In view of the real-time requirement of road segment identification,the discriminant model of road traffic state of rough sets are based on Fuzzy Clustering.The model mapping through both the traffic impact index and traffic state,integrate the influence of different traffic conditions of the impact of the index set,to generate a more streamlined traffic state judgment rules.An example of this is provided within Chapter 4 and Chapter 5.(4)In order to make the identification,the results can take into account the network characteristics.This paper proposes a traffic network identification model based on multi-level fuzzy comprehensive evaluation.When the method is used to evaluate the road traffic state,the imbalance of the static characteristics of the road network and the influence of the dynamic traffic characteristics on the indexes are considered.Both the subjective and objective factors are considered to optimize the weight distribution of each factor in the comprehensive evaluation.The results show that the model proposed in this paper provides a higher accuracy rate when it is utilized to judge the traffic state of a road network.