Research On Travel Time Estimation And Short-term Prediction Of Urban Road Based On Automatic Number Plate Recognition Data

With the aggravation of urban traffic congestion,providing real-time and accurate traffic information for road users and traffic managers has become one of the key challenges for intelligent transportation systems.Compared with other traffic parameters,travel time can more intuitively reflect the real traffic state of the road,including traffic congestion and special events.Therefore,the research of travel time has become an important content in the field of modern and future transportation science and engineering.As an open traffic network,vehicles will be affected by many uncertain factors,such as random delay,traffic signal control,traffic demand fluctuation,etc.The traffic flow of urban road presents the characteristics of interrupted distribution,and the randomness and internal uncertainty of travel time distribution are strong.Therefore,this paper studies the interrupted travel time of urban roads based on automatic license plate recognition data.The main research contents are as follows:(1)Based on the analysis of the problems and errors in the original data,a complete data preprocessing process and quality analysis method are proposed.At the same time,this paper introduces the basic process of obtaining link travel time by license plate matching,and makes necessary prior analysis and pre-cleaning for the original travel time data set,so as to provide a complete and reliable basic data set for the follow-up travel time estimation and short-term prediction research.(2)Based on the analysis of the travel time characteristics the interrupted flow of urban roads,a travel time estimation algorithm suitable for different traffic states is proposed.According to different driving characteristics,the algorithm can divide the traffic flow into non signal delay component,signal delay component and short-term activity component.By analyzing the vehicle characteristics in different time periods and different driving states,it is found that short-term active vehicles are the important cause of road traffic congestion.Only by eliminating the noise data generated by short-term active vehicles can accurate travel time data be obtained Set.(3)This paper analyzes the temporal and spatial correlation of traffic flow data,innovatively introduces the standard deviation of relative speed as a part of the eigenvector,and uses the Gaussian mixture model to fine segment the historical database according to different traffic conditions,providing data support for the short-term travel time prediction.(4)At the same time,four machine learning algorithms are used to predict the travel time,which verifies the robustness of support vector machine to the prediction of finite sample set.At last,the empirical mode decomposition is proposed to correct the error in order to optimize the prediction results.