Simulation of traffic incident detection based on VISSIM and neural network

In recent years, the construction of highway in China has achieved remarkable success; meanwhile, the unbalance between the transportation demand and capacity causes traffic incidents continuously. To develop a high efficiency and reliable incident detection system is the most important way to solve the problem. However, a good incident detection system relies on its detection algorithms. In this paper, we use VISSIM to simulate an incident that caused by a car suddenly stop on it, and then collect the parameters before and after the incident. After obtaining the original simulated data, we use DB2 wavelet analysis to process the traffic parameters. Then we propose the data in different kinds of neural network algorithms: BP, SOM and RBF neural network, in order to update a new method with high detection rate, low false alarm rate and short detection time.