关键词: 事件检测, 神经网络, 交通流

Abstract: Directing at the problem of mode identification of incident detection in transportation field，a novel method is proposed for traffic incidents detection based on improved adaboost method.The structure and training algorithm of Kohonen neural network are formulated.Then the influence of an incident on the traffic flow is analyzed，and Kohonen neural network input variables are selected reasonably.At last Adaboost algorithm is used to build an integration-neural network.Finally we simulate with Matlab and find out the algorithm has such advantages as fast learning speed，good generalization ability and high detection rate.

Key words: incident detection, neural network, traffic flow