Research On Fusion And Prediction Methods Of Highway Multi-source Heterogeneous Traffic Data

With the rapid development of Chinese economy, increase in population,incomegrowth, increase in numbers of vehicles, which make travel proportion of vehicles isincreasing and urban traffic is facing enormous challenge. Traffic congestion, trafficpollution, traffic safety problems occur frequently, which have a negative impact inpeople’s daily lives, and also challenge the sustainable and healthy development ofChinese economy. In china traffic datas is short, which leads to the problem of trafficmanagement, planning and control and other aspects in many cities. Trafficmanagement requires lots of traffic datas, but the traffic datass obtained by a singledetector do not meet the needs of the traffic datas obviously. With the emergence ofthe information fusion technology and have developed rapidly, this problem can beresolved effectively. In accordance to the above-mentioned problems, the maincontents of this paper are multi-source heterogeneous traffic datas fusion andprediction methods. First multi-source traffic information quality evaluation andcontrol technologies are researched in highways. we propose two traffic datasprocessing methods,of which are fixed collection and moving collection fusionmethod, traffic flow forecasting methods based on historical datas and real-time datas.Fixed collection and moving collection rapid fusion method. This paperintroduces the basic principle of the minimum variance weighted average method anddetermining method of weight. Because the current traffic information collectionmethods main are video devices and floating car, so this paper describes an exampleof interval average speed. Finally, the fusion method function module is designed andthe proposed method is verified. Fixed collection and moving collection rapid fusionmethod proposed in this paper proved is simple and practical by the experiment trafficflow forecasting methods based on historical datas and real-time datas.Traffic flow forecasting methods based on historical datas and real-time datas.Due to technical and management limitation, of, detectors are often destroyed or somesections do not install detectors in highways, which result in some sections can notdetect traffic datass, and sometimes detectors detects traffic parameters are inaccurate because of the weather problems. First this paper analyze historical datass of notinstalling dectors and adjacent sections using cluster analysis, then real-time andhistorical datass are fused. The historical datass of upstream sections and floating carof not installing dectors sections datass are fused using traffic flowing theory. Finally,the fusion function module of historical datass and real-time datass is designed and itis verified by experiment.