Research On The Method Of Recognizing Traffic Mode Based On GPS Positioning Data Of Mobile Phone

Modern urban traffic behavior modeling and urban traffic problem analysis rely on larger sample size and more detailed individual travel survey data.Traditional resident travel survey method has many defects,such as low data quality,cumbersome survey organization,long data update cycle and so on,which affect the quality of individual travel data collection for a long time.The traffic planning and management schemes formulated based on these data are also difficult to obtain the expected effect.Therefore,how to explore a more objective,accurate and efficient individual travel information collection method has become an important problem throughout the traffic engineering field.In recent years,with the development of mobile positioning technology,mobile GPS can continuously collect individual travel space-time trajectory information,help to restore and reflect individual travel characteristics in the whole process,brings an unprecedented opportunity for refined individual travel information collection.It is of great significance to study the traffic information extraction method based on mobile phone GPS data.In this thesis,a multi-mode travel behavior information recognition method based on mobile GPS positioning data is proposed.First,according to the actual travel characteristics and habits of Chinese residents,comprehensive and diverse multi-mode combined travel experiments are set up in different periods and road sections.Travel modes include walk,bicycle,bus,car and railway etc.The characteristics of smartphone GPS data of different trip modes are analyzed in detail,and the representative indicators of different modes are extracted.Second,the methods and algorithms of individual trip mode identification by using smartphone GPS data are proposed.On the one hand,this thesis constructs and applies the BP neural network model to identify trip modes of walking,bicycle and motor vehicle,and explores the influence of different model parameters and input attributes.On the other hand,the bus stop matching algorithm is designed to further identify the bus and car in the motor mode,which successfully solves the difficult problem of distinguishing the bus and car.The recognition accuracy of all travel modes are more than 85%.Finally,this thesis objectively analyzes and evaluates the technical feasibility and effectiveness under different sampling frequencies and traffic states,and gives some suggestions for the popularization and application of the proposed method.This study deeply excavates the residents’ travel characteristic data collected by mobile GPS sensors,and successfully carries out neural network model based pattern recognition for the combination of multi-mode travel modes.This research can provide effective data support for urban traffic demand prediction and traffic model evolution.