Research And Implementation Of Urban Traffic Condition Sensing And Prediction System By Fusing Multi-source Data

The advancement of urbanization has resulted in more frequent traffic congestion and traffic accident,which makes most cities in a sub-health state of traffic,and leads to extra time cost,and even loss of life and property.However,the existing traffic information systems lack real-time sensing of traffic data and intelligent regulation of traffic condition.Consequently,in this thesis,we design an urban traffic condition sensing and prediction system based on multi-source data to sense and predict traffic conditions in real time.The main contribution of this thesis is as follows:Firstly,we design a multi-source traffic data aggregation and calculation platform using data stream processing technology,which is capable of fusing multisource heterogeneous data such as traffic card crossing records,taxi trajectories and accident data to sense traffic condition,including intersection flow sensing,road condition sensing and spatial and temporal analysis of congestion modules.Secondly,we design the traffic flow forecasting module and accident risk prediction module based on the real-time sensing data.When it comes to traffic flow prediction,in view of the lack of semantic information embedding in the existing methods,we propose a model based on multi-graph convolution and recurrent neural network to fuse the semantic features for traffic flow prediction.As for accident risk prediction,in order to address the sparsity and spatial heterogeneity of accident data,we propose an accident risk prediction model based on the combination of scale reduction attention mechanism and graph convolutional neural network,where a downscaling module is leveraged to help predicting the road-level accident risk.Finally,the results on two real-world traffic datasets show that our method outperforms other baseline methods.This system can sense the traffic congestion and capture the largetraffic intersections in advance,and thus assist traffic guidance and traffic police work.