The Research Of Traffic Flow Detection Based On Information Fusion

Traffic flow detection is an important component part of ITS, which can obtain the accurate and real-time data of traffic flow from all sensors. Then the goal of intelligent control could be achieved by sending information to Intelligent Traffic Control Centre and send out corresponding order.Traffic flow detection technology has developed from a single sensor processing system to a multi-sensor system. The various types of sensors can be optimization set and complimented mutually. By using the data fusion of multi-sensor, the current research of traffic flow data processing technology is focused on how to improve the accuracy of data processing and reliability effectively through valid data-processing algorithms. To detect traffic flow parameters, algorithms are applied recently, such as neural network algorithm, Bayes decision theory, expert systems, genetic algorithms, Kalman filtering etc, which are all achieved certain achievements. The traffic flow detection method based on information fusion is proposed to detect and predict the traffic flow parameters.Compared with traditional method, this method is not meant to apply single data detection and prediction technique or one message, or to fit these together simply, but meant to widen data, to optimize detection way, and to fuse data and method soundly. On the one hand, it can evidently increase detection and prediction accuracy; on the other hand, robustness of system is markedly enhanced.With an example of ring coil detector, the traffic flow detection method based on data fusion is proposed. And this paper is composed by following:(1) In the process of the velocity detection, the detection method of distribution and the method of removing failure data based on compatibility matrix are applied to detect the data validity for improving the accuracy of data detection. The principle of velocity detection method based on the weighted average is explained. The advantages and disadvantages of Weighted Average of Minimum Square Error (MSE) and Adptive Weighted Average are analyzed. By simulation, the precision of velocity detection method based on the weighted average is testified.(2) Base on the requestion of estimating distribution of traffic flow by using Bayes theory, the model of traffic flow is established, according to which the algorithm of detecting the distribute of traffic flow based on Bayes parameter estimation theory and the algorithm of detecting the distribute of traffic flow based on Bayes risk decision theory are adopted. By analyzing two algorithms, the characteristic of Bayes risk decision theory in the process of trsffic detection is sound when the risk coefficients are different. The reasonability of decision is proved.(3) For the purpose of immunity to outliers and improving the precision of prediction, Kalman filtering prediction algorithm of immunity to outliers based on M estimation theory is brought forward on the foundation that applying Kalman filter theory to predict the queue length. The better robust is illuminated by simulation.The mentioned method and algorithm are testified valid on the basis of simulation-generated flow data and velocity data.