Study On The Method Of Signal Control Intersection Traffic Congestion Identification

Traffic congestion has attracted the attention of all countries in the world and became the problems to be solved. Traffic energy consumption, environmental pollution caused by traffic congestion, is one of the most serious "Urban Diseases" in our country. Traffic congestion identification can prevent and alleviate the city traffic jam, so it is of great significance to study. According to the characteristics of urban road traffic congestion, combined with the characteristics of signal control intersection, this thesis mainly discusses the method of the signal control intersection congestion identification based on Bayesian decision, the method to update the Bias training sample set, and the design of the system of signal control intersection traffic congestion state recognition based on naive Bayesian.Research on congestion identification method of intersection signal control, with Bias theory and simple Bias classifier model as the basis, the thesis puts forward a method of signal control intersection congestion identification based on simple Bias decision, it considers the recognition of traffic congestion as a classification problem of uncertainty, the state is divided into smooth, crowded and congestion, and it uses traffic flow, occupancy rate and queue length rate as the discriminant parameters, generate Bias classifier by learning historical data of smooth, crowded and congestion three conditions, and then use the classifier to classify the real-time acquisition data, completing the identification of traffic state.The Bias classifier model is based on the probability table of the history training sample, so the best training sample can naturally decide the classification tendency of the classifier. Preparing training samples is the basic work of Bayesian classification, but in practice it is difficult to obtain sufficient training samples and it is not comprehensive to identify the traffic state reliing solely on the static historical data. Based on this, this thesis proposes an incremental learning method to update traffic training samples and an improved recognition algorithm, which adds classified data to the training set with certain rules, dynamic updates of the training set, riches training information, making the results of congestion identification more reliable.This thesis analysis and evaluates the Bias algorithm by the VISSIM simulation data, the error rate of algorithm is 6.92%, showing that the algorithm is feasible and practical for the congestion identification of signal control intersection. Finally, guiding practice with theory, it introduces the overall architecture of signal control intersection traffic state identification system, constructs the system of control signal intersection traffic state recognition based on the Bayesian method, including functional design and database design. In the integrated development environment of C# programming language, using the structured design, it uses softwear to realize the functions discussed, to identify the traffic state.