Researh Of Freeway Traffic Incident Detection Based On PNN Optimised By Chaos And GA

In recent years, the rapid development of highway to meet the increasing traffic demand, bring great convenience to the people, and promote social development. With the increasing of vehicle ownership in China, at the same time of it bring convenience, traffic congestion, traffic accidents caused by them have also been widespread concern from all walks of life. Once the event (incident) occurred on the expressway, caused by the loss is difficult to predict. Therefore, it is necessary through the study of freeway traffic incident detection, improve the efficiency of highway transportation, to protect people’s life and property and provide fast and comfortable transport road environment.Based on traffic flow theory elaborated, the paper analyzes the traffic parameters change in case of traffic incident situation. In order to better highlight the events before and after the traffic conditions, we choice the combination of traffic volume, average speed as the input feature vectors of detection model. Then, Simulation analysis by I-880 traffic database, in order to verify the feasibility of the model.The probabilistic neural network (PNN) for freeway traffic incident detection. The paper simulate the model with the real traffic data. The PNN model smoothing factor is difficult to determine, the shortcomings of the model of the structure of the redundancy, this paper by means of real number coding, adaptive crossover and mutation to improved genetic algorithm (GA), forming a synthetically improved genetic algorithm (IGA) and which can be used for optimizing the PNN model smoothing factor and pattern layer structure, then construct IGA-PNN incident detection model. Through the simulation found the model have higher detection rate and accurate rate. Finally, the introduction of chaos (Chaos) concept. Initialization population genetic algorithm by chaotic method. The evolutionary process of genetic algorithm, using chaotic search parameter space near the best individual. Through the method of chaos, increase the diversity of the initial population, improve the ability of local search algorithm. Then, based on the CGA-PNN traffic incident detection simulation, validates the method compared to BP, IGA-PNN method has higher detection rate and high accuracy. For the freeway traffic incident detection gives the new method.