Research On Freeway Traffic Condition Monitoring Based On Vehicle Trajectory Data

As an important component of essential transportation infrastructure,freeways are undergoing a phase of intelligent development with the support of national policies and the advancement of intelligent,information,and digital technologies.The intelligent development has brought about a significant transformation in the management mode of freeways.In the past,traffic managers could only passively respond to congestion or other traffic issues without prior knowledge or measures.However,through the application of intelligent technologies,traffic managers can now promptly obtain comprehensive and accurate traffic operation information,enabling them to make informed decisions swiftly.This shift towards proactive management helps improve the operational efficiency and safety of freeways,providing users with better travel experiences.Therefore,understanding the operational status of traffic flow becomes a crucial aspect of proactive management,and this paper focuses on the monitoring of traffic conditions using data obtained from road detectors.The quality of detection data highly influences traffic condition monitoring.Traditional detectors providing aggregated data fail to meet the requirements of realtime and highly accurate monitoring,as they suffer from low detection accuracy,time delay,and narrow coverage.Vehicle trajectory data provides more detailed information for monitoring the traffic status on freeways,and extracting traffic parameters that can finely characterize the operational state of traffic flow becomes a current research focus and challenge.To address this issue,this paper incorporates correlation analysis and entropy weighting to select feature parameters that can effectively represent traffic conditions.Furthermore,the fuzzy C-means algorithm(FCM)and random forest(RF)classification algorithm are introduced to establish an FCM-RF traffic state discrimination model.The SMOTE algorithm is employed to address the issue of insufficient anomaly state data samples by balancing the data samples.This article conducts an instance analysis and validation based on the monitoring data of the Ganzhou-Anhui border to Wuyuan section of the Dezhou to Shangrao freeway.Firstly,through data cleaning,transformation,integration,and dimensionality reduction,denoised vehicle trajectory data is obtained.Then,considering the strong correlation among traffic parameters,a traffic parameter extraction method combining correlation analysis and entropy weight method is proposed.Among numerous parameters,four traffic parameters,namely speed,speed difference,headway distance,and acceleration,which can accurately represent traffic states,are determined.This achieves dimensionality reduction of traffic state representation parameters and reduces the computational complexity of the traffic state discrimination model.Secondly,through comparative analysis of domestic and international traffic state classification standards,it is found that the current standards mainly characterize traffic states based on macroscopic parameters,without analyzing traffic states at the microscopic level.This leads to inaccurate discrimination of traffic states and inability to provide effective information for proactive management and control.Therefore,this article takes micro-level parameters such as speed,speed difference,headway distance,and acceleration as indicators to construct a traffic state classification method based on fuzzy C-means clustering algorithm,which can simultaneously evaluate traffic efficiency and safety.Furthermore,based on the proposed traffic state classification method,the fuzzy C-means clustering algorithm is used to perform clustering analysis on the traffic flow dataset,yielding traffic flow sample datasets of different states.To address the problem of low classification accuracy caused by imbalanced datasets,the SMOTE algorithm is used for data balancing.By establishing the FCM-SMOTE-RF traffic state discrimination model and using the confusion matrix and precision as performance evaluation indicators,the results show that the accuracy of traffic state discrimination is above 94% and can achieve accurate identification of traffic states.