Identification And Prediction Of Highway Tunnel Traffic Operation Status Based On Spatial And Temporal Characteristics

With the increasing conflict between traffic supply and demand at key access points such as highway entrances and tunnels,it has become an urgent engineering problem to achieve intelligent control of highway traffic operation and improve its transportation efficiency.In this paper,we obtain video information through tunnel side cameras and obtain traffic operation data based on target detection algorithm.On this basis,the traffic operation characteristics of the highway tunnel are analyzed in depth,and the identification and prediction of traffic operation status are carried out from the perspective of time and space respectively.Firstly,a deep learning framework based on CSPDarknet-53 network is built,and the YOLO v7 algorithm is used to parse the tunnel monitoring video and obtain the traffic operation data such as vehicle speed,workshop distance and traffic flow.In the same time and space dimensions,the spatio-temporal scenarios are divided into time periods,lanes and zones,and the traffic operation characteristics of highway tunnels in different spatio-temporal scenarios are analyzed in depth.In the temporal dimension,the traffic flow and average travel speed show "M" and "W" type trends respectively,showing distinct similarity and periodicity,with obvious time-varying characteristics.In the spatial dimension,the speed variation of different lanes and sections in the tunnel is obvious,and the average travel speed variation of different lanes can reach 21.5%,and the average travel speed variation of different sections reaches 9.1%,with significant spatial variability.Secondly,the average travel speed and traffic flow were selected as traffic operation status identification indexes,and the operation status was classified into four types: smooth,smooth,congested and jammed.The GBK-means clustering algorithm is used to classify the traffic operation states and obtain the clustering centers of different traffic operation states.On this basis,the traffic operation states are further identified using the support vector machine(SVM)model,and the spatial and temporal characteristics of the traffic operation states are analyzed.The results show that the clustering effect of GBK-means clustering algorithm is better than that of traditional K-means clustering algorithm,and can better characterize the traffic operation of highway tunnels;the accuracy of traffic operation status discrimination using SVM model reaches more than 80%,and the reliability of the identification results is high.Finally,a time series prediction model based on TSK fuzzy neural network structure is proposed to predict the traffic operation characteristics parameters,and the applicability and accuracy of the model are evaluated by the error sum of squares,coefficient of determination,average absolute percentage error and error variance,and the model is validated and analyzed by using actual data.The prediction results show that the evaluation indexes of the model indicate that the model has a high degree of adaptability to the traffic operation data and can accurately predict the trend of the traffic characteristic parameter data.