Research On Key Technologies Of Urban Traffic Trajectory Data Mining

With the continuous advancement of urbanization in China,the scale of cities has gradually increased,and urban development has been characterized by diversity,dynamics,and complexity.It has placed higher demands on urban scientific governance,and data mining has become a supporting technology,especially with the construction and development of smart cities.It is urgent to mine knowledge from the massive urban operational data to support the application of various industries.In recent years,the popularization of mobile positioning technology and the development of wireless communication technology have accumulated a large amount of mobile trajectory data in the urban transportation field.It is one of the important research objects of smart cities,and has become a new type of data other than image,video,audio and other media data.Massive mobile trajectory data has important social and economic value.While recording urban traffic conditions in real time,it also reflects urban residents' travel rules and social spatial structure.How to mine hidden knowledge from trajectory data has become an important research topic.The urban traffic trajectory data is a record formed by the spatial position of a moving object such as a vehicle in time in a city,and mainly contains semantic information such as time and space,and is usually represented by a time series composed of time-stamped position coordinate points.Trajectory data mining generally includes key technologies such as data preprocessing,indexing and querying,behavior pattern mining,and trajectory visualization.The research scope can cover the data itself,urban traffic status,and urban residents' travel activities.This thesis takes urban traffic trajectory data as the research object,and studies the problem of moving track data,such as stay point extraction,trajectory-directed line query and index,urban travel hotspot mining,an so on.The main work and innovations are as follows:(1)The thesis proposes an interactive moving track stay point extraction method.The stay points contain important semantics.Extracting the stay points is the basic step of many trajectory data mining.The current method of staying point extraction generally requires pre-setting threshold parameters,and the parameters have a great influence on the extraction results.Most of these parameters rely on empirical settings.There are various types of moving objects,and the trajectory data varies greatly due to different geometric physical properties and driving environment.It is not easy for non-professionals to set appropriate empirical parameters.Aiming at this problem,this paper designs an interactive visualization method to extract the moving track stay point.Firstly,the Space-Time Cube(STC)method is used to visualize the moving track,then the user interactively selects the typical trajectory stay points.The geometrical features of the regional 3-dimensional bounding box automatically determine the threshold parameters,and finally the sliding window method is used to extract the staying points.The method helps to quickly and intuitively determine the threshold parameters,and can effectively improve the efficiency of the non-professionals to extract the trajectory stay points.(2)The thesis proposes a trajectory-directed line similarity quantitative calculation method for sketch retrieval.Similarity calculation is the core problem of trajectory data mining.The key to trajectory query,clustering and anomaly detection is the similarity calculation.In urban traffic,moving objects may move in,out,cross,and stay multiple times when moving relative to a directed line(such as a road).There are currently few studies on trajectory-directed line topology queries.Aiming at this problem,a sketched trajectory-directed line query mode is designed,and a trajectory-directed line similarity quantitative calculation method is proposed.The sketched stroke-directed line is used as an input query condition and is described as a sequence of key points for semantic association.The trajectory-directed line to be queried is also described as a sequence of key points.The similarity between the two is calculated by the point sequence distance metric.For the trajectory-directed line with different number of key points,two kinds of distance matching methods are used to verify the similarity.The results show that the method is simple and feasible,and can realize quantitative query of trajectory data.(3)The thesis proposes a trajectory index and query method for road network.The moving object index for urban traffic road network is an important part of trajectory data management.The current research on trajectory data index mostly focuses on the index and query of moving object position,and does not support trajectory-directed topology query.For the trajectory-directed line query problem,a two-layer index structure ISTR-tree is designed.The upper layer uses R*-tree to index the road network,and the lower layer uses R-tree to index the trajectory-directed line key points.Calculating the DTW distance lower bound can speed up the query process.(4)The thesis proposes a method of urban travel hotspot mining based on trajectory data.Mining urban hotspots and their interactions is an important means to explore residents' travel patterns and mobile patterns.Aiming at the lack of data aggregation verification in current hotspot extraction methods,we propose a travel hotspot mining method based on taxi trajectory.Firstly,the taxi pick-up and drop-down points are extracted by geometric and time constraints,and then the spatial autocorrelation test is performed on the number of passengers and passengers in the grid through the meshing method to verify the clustering of the trajectory data,and then the taxi pick-up and drop-down points are gathered.Into the hotspot area,and establish a hotspot interactive network,quantitative analysis of hotspots and their interactions,the experimental results show the effectiveness of the method in this chapter.