Design And Implementation Of Traffic Area Division System Based On Spatial-Temporal Trajectory Mining

With the advancement of urbanization and the continuous increase number of private vehicles,urban traffic problems caused by traffic congestion have become an important bottleneck restricting socio-economic development and the improvement of people's living standards.As a common strategy to alleviate traffic congestion,traffic sub-area division divides the entire traffic network into several different sub-areas to facilitate traffic analysis at a finer-grained level and hence reduce the-corresponding complexityExisting related research mainly focuses on spatial division,and ignores the distribution changes of sub-areas over time,accounting for it is impossible to carry out traffic network division with more accuracy.Given this,this thesis designs and implements a traffic area division system based on spatio-temporal trajectory mining,by mining and analyzing trajectory data with spatio-temporal semantics,identifying time intervals with densely distributed trajectories,and then detecting the spatial heterogeneity within each time period,and carrying out boundary recognition through convex hull algorithm to achieve the corresponding sub-area division within different time intervals to meet the time-varying needs of traffic control.Experiments show that the system can better identify the traffic sub-areas and their corresponding time intervals compared with the existing traffic sub-area division methods.The main research contents of this thesis are as follows:(1)Research on data preprocessing methods based on trajectory characteristics,and create a data preprocessing model.It first uses mean filters and fast sorting algorithms to repair trajectory drift points and remove redundant data.Then,based on the characteristics of trajectory retention points,it introduces the spatial-temporal neighborhood to improve the DBSCAN algorithm to detect and retain trajectory retention points and remove worthless information.Then,combined with line segment simplification and angle offset,a trajectory segmentation compression algorithm based on double offset limitation is proposed to identify feature points and come out segmentation compression to optimize trajectory quality,which finally provides concise and effective data support for the subsequent spatio-temporal trajectory mining and analysis.(2)Research on the key technology of traffic area division based on spatiotemporal trajectory mining.This thesis firstly extends the classic partitioning algorithm K-means,and applies it to the time dimension to extract intensive time intervals where vehicles frequently occur,so as to convert the entire transportation network partition problem into a single intensive time interval partition problem.Moreover,in view of the shortcomings of the Density Peaks Clustering algorithm,the k-nearest neighbor concept is introduced to redefine the concept of density,and a clustering center discrimination filtering mechanism is proposed to reconstruct the clustering model.The spatial heterogeneity of areas is found in the semantics implicit in the trajectory,consequently the regions are divided(3)Design and implement a traffic area division system based on spatio-temporal trajectory mining.The system integrates trajectory data processing,spatio-temporal trajectory clustering analysis,and traffic area detection and identification.It can confirm a set of traffic sub-areas and their corresponding time intervals without user intervention for the purpose of urban planning.