The Method And System Implementation Of Traffic Flow Operation Feature Extraction Based On Bayonet Monitoring Data

Urban transportation has entered a new era of information,intelligence and wisdom.Combining big data mining,machine learning and other technologies with intelligent transportation can significantly improve the traffic congestion in urban development.With the increase of video monitoring equipment,such as electric police and bayonet,the traffic database has been greatly enriched,which provides driving force for intelligent and fine traffic management.However,the traditional traffic data processing method can not adapt to the rapid growth of data.The traffic information behind the traffic bayonet data needs unified process dealing with framework and application methods to analyze and practice.Aiming at the application problems of checkpoint data,this paper combined with the basic information of road network to fully tap the potential of checkpoint data,and carried out methodological research in many application directions such as traffic flow parameter extraction,trajectory data clustering analysis and traffic flow prediction,According to its research methods,Python and web technology were used to build application system,which could solve the actual traffic problems.The main research contents are as follows.(1)Traffic flow state parameter extraction based on Data Mining.Based on Python and ecarts,this paper proposed a complete set of traffic checkpoint data mining process framework,which included preprocessing,index parameter extraction and visualization.The framework used real-time extraction of delay,flow,speed and other parameters to achieve active full-time state extraction.Then,based on the extraction of travel time,average speed and vehicle cycle delay,the queuing length estimation model was constructed by using the number of queuing vehicles.Finally,taking the trunk road of Beijing Road in Zhangdian District of Zibo City as an example,the automatic extraction and visualization of traffic operation parameters were realized,and the queue length was selected for precision analysis.The accuracy of the model could provide support for the decision-making of intelligent traffic control.(2)Analysis of spatial and temporal characteristics of traffic flow based on trajectory clustering.The threshold value was set by the travel time and speed factors extracted by traffic parameters.The road network topology and TOPSIS method were used to complete the extraction and complement of vehicle travel chain.Then,LCSS algorithm was used to measure the similarity of trajectory chain,and the cluster processing of trajectory data was carried out by DBSCAN density clustering algorithm.Based on the clustering results,the spatial and temporal characteristics of vehicle travel were studied.Finally,taking the west area of Zhangdian District of Zibo as an example,the track extraction and cluster analysis were carried out.The experimental data were divided into four groups of sample data:working day morning,evening peak,rest day morning and evening peak.The spatial and temporal travel characteristics obtained by clustering the sample data were in line with the actual travel situation.(3)Short term traffic flow prediction based on operation situation.In this paper,the hidden Markov model was used to predict the traffic flow potential.The LSTM neural network was trained together with the cycle average speed and historical flow data.Besides,the multipul and multi-step prediction method was used to improve the prediction accuracy and reduce the prediction lag.Finally,taking Shiji road in Zhangdian District of Zibo City as an example,the performance of the improved model was tested.The experimental results showed that the accuracy of the improved LSTM traffic flow prediction model was better than that of the traditional LSTM model.(4)Design and implementation of traffic flow operation feature extraction system.Based on the framework of traffic flow parameters extraction,the system used the method of trajectory data extraction and clustering.The improved traffic flow prediction model could design the function module.On the premise of using its algorithm logic process,it selected the mature background framework Spring Boot and front-end framework React to develope the system.Finally,according to the system design principles,the function of the submodule was refined and designed.Thus the traffic flow feature extraction system has been built completely.