Research On Traffic Congestion Analysis Based On Large Trajectory Data

With the rapid development of social economy,the number of vehicles in the city increases rapidly.Thus,the traffic pressure is also increasing.At present,the existing traffic carrying capacity and operation capacity can no longer meet people's ever-expanding needs for scarce land resources and traffic resources,which has brought serious urban diseases.Traffic congestion not only hinders people's travel and aggravates social conflicts,but also causes air pollution in cities.Besides,traffic congestion also leads to the increase of accidents,which in turn aggravates the traffic congestion.Moreover,the economic losses caused by traffic congestion are huge every year.How to solve urban traffic problems has become the focus of the smart city research.As the rapid development of the advancedinformation technologies,trajectory data is growing rapidly as well.How to mine the useful information and find out the causes of traffic congestionfromthe trajectory data,which have become an urgent problem to be solved in current.In this paper,by studying the characteristics of traffic trajectory data,considering with the clustering algorithm and the trajectory data clustering algorithm,a novel congestion point computation method is proposed,which can efficiently identify the urban traffic congestion points The main work of the paper are listed as follows:(1)A clustering cluster number discovery algorithm based on grid centroid is proposed,which can effectively discover the cluster number of a data set.The current k-means clustering algorithms(including the improved k-means algorithms)have a lot of shortcomings,which requires the determination of the number of clustering clusters and the random selection of initial center points.At the same time,compared with the existing methods,not only the time efficiency of this method is also greatly improved,,but also it provides an efficient method for k-means to select the initial point.(2)A grid-based K-means parallel clustering algorithm is proposed.Compared with the existing k-means clustering algorithm and most of the improved algorithms based on k-means,this algorithm has significantly improved the execution efficiency of the algorithm on the basis of ensuring accuracy error.(3)Through the analysis and research of urban road network and trajectory data,an urban road network trajectory clustering algorithm based on the maximum density of grid is proposed.The algorithm can cluster the entire urban road network structure and divide the vehicle trajectory density into different zones,so as to find the hot areas of urban traffic.On this basis,a new method of urban traffic congestion detection is proposed.Unlike most traffic congestion detection methods,this method is no only limited to finding congestion areas,but also can find the geographical location of the origin of urban traffic congestion.At the same time,a congestion identification method for traffic signal intersections is proposed based on the special traffic models such as traffic signal intersections and intersections.