Research On Travel Speed Prediction Based On Traffic Data Fusion

This paper mainly focuses on the analysis and research of traffic data fusion in the field of intelligent transportation system.On the one hand,the speed of travel is one of the hard fingers reflecting the state of traffic.In the actual traffic condition,the time and space change rule of the traffic data flow is time-varying because the volume of traffic is undulating in the day,which indirectly results in the change of the travel speed depending on the change of traffic flow.Therefore,accurate travel speed prediction is an important basic data reflecting ITS services.On the other hand,due to the high uncertainty,randomness and complexity of the traffic data on the travel section,the effect of the single type detector based on induction coil and GPS floating car is not so satisfactory.In view of the complementarity of the two detection techniques in terms of prediction speed,based on the multisource traffic data fusion,this paper proposes an improved BP neural network algorithm to fuse the travel speed predicted by the two detectors in order to further improve the accuracy of the travel speed.First,the traffic data obtained by the induction coil detector and the GPS floating car detector are identified by the threshold method and the traffic flow theory,and corrected this data.In this paper,the mean value method of the adjacent time period is proposed to improve the quality of the collected data.Then the principles,concepts,levels and methods of data fusion are introduced,and the application of data fusion technology in the advanced traffic information system is studied on the macro level.For different single type of detectors in predicting travel speed.In this paper,on the prediction of the travel speed of the data fusion algorithm for multiple floating car,which introduces the basic weight value of the single sample of the floating car,the weight of the processing strategy,the weight value of the road condition state and the fusion coefficient corresponding to the three,through the speed of travel is predicted by the fusion algorithm.Finally,due to the inaccuracy of single type detector in calculating the speed of travel,an improved BP neural network is proposed to fuse the travel speed data obtained by two detectors,and a travel speed prediction model based on traffic data fusion is established.