Congested Detection Of Freeway Section Based On Multi-source Data Considering Distribution Of Detective Equipment

Congestion detection of expressway sections is the necessary prerequisite for effective management and timely control.Due to the limitation of data conditions,most of the data studied in the past come from a single data source,which can not represent the comprehensive traffic state from many aspects.At the same time,due to the influence of installation distance of testing equipment,the existing testing models can not meet their respective applicable conditions,which leads to the difficulty of improving the detective rate.At present,the layout and popularization of detection equipment in expressway provide conditions for acquiring multi-source data.Considering the distribution of detection equipment,it is important for improving the quality of expressway management to make rational use of existing expressway data sources and establish an effective road congestion detection model.Therefor,the research combines the vehicle detector data,video images,toll data and bayonet data of expressway,aiming at the situation of dense and sparse distribution of detection equipment.making full use of the traffic parameters provided by various data sources to realize the road congestion detection by fusion method,and establishing the road congestion detection model of expressway to improve the detection rate are of great significance for improving the quality of expressway management.The main research contents include:(1)In view of the dense distribution of detection equipment,the decision-level fusion method is adopted to detect the congestion in the road section.The average travel time of current expressway sections is estimated based on toll data and bayonet data to reduce the impact caused by the delay of travel time.Then,a congestion detection model is established based on the data of upstream and downstream vehicle detectors.Finally,considering the condition of dense equipment distribution,the video detection results and the above two results are fused at decision level to established a congestion detection model.The result shows that when distribution of detection equipment is dense,the detection model based on decision-level fusion has a good result.(2)In view of the sparse distribution of detection equipment,the feature-level fusion method is adopted to detect the congestion in the road section.firstly,based on the data of upstream and downstream vehicle detectorss,toll collection and bayonet,the relevant features are extracted.Realizing feature-level fusion to established congestion detection model by combining unsupervised learning with supervised learning.The result shows that when distribution of detection equipment is sparse,the detection model based on feature-level fusion has a good result.(3)In order to verify the pertinence and applicability of the detection model,the decision-level fusion method and feature-level fusion method are used to compare the detection results under the condition of dense and sparse distribution of detection equipment.The result shows that when distribution of detection equipment is dense,using decision-level fusion method is better,and when distribution of detection equipment is sparse,using feature-level fusion method is better.On this basis,a one-month congestion detection application on the road section shows that the two models proposed in this paper are feasible and the detection effect is good.