The Application Of Neural Network In Urban Traffic Flow Data Fusion

As a new subject, data fusion has developed rapidly in military surveillance and defense system, robotics, intelligent traffic, environmental monitoring, object identification and medical diagnosis since 1980s.Based on the traffic flow data fusion, the application of the artificial neural network in this area is researched and discussed in this paper.First, a brief introduction of the concepts and models of data fusion is given, and then, some algorithms of data fusion are provided. This paper presents the advantage of applying the neural network into data fusion and make sure that the neural network is major objective for research.When there are many detectors, cluster analysis should be first applied to cluster lanes include single point controlled intersections without detector and single controlled insections with detector, and then predict the traffic flow in the single point controlled intersections without detector by the use of the relationship between these two. The aim is to minimize the number of variables in the predictable model in order to maintain the stability of the modle.At the same time; these variables almost contain information of all variables in certain class so as to enhance prediction accuracy.Finally the paper introduces the basic concepts and models of neural network, as well as the basic methodology and principles of neural network using in the model of data fusion. A method of data fusion is constructed using neural network, which is applied to the traffic control system, combined the simulant data by VB, and verifies the validity of the method by training and simulating of neural network. The results indicate the method to improve traffic flow of detection is valid.