Design And Implementation Of Hotspot Area Mining System For Road Network Based On Mobile Spatio-Temporal Trajectory

In recent years,with the rapid development of wireless positioning technology and the increasing number of motor vehicles,the acquisition of moving object trajectories has become more and more convenient while the number of trajectories is exploding.These data contain the spatio-temporal dynamics of objects and their behavioral characteristics.Therefore,how to effectively deal with these massive trajectory data and obtain knowledge has become a hot research topic.Especially in the transportation field,large-scale vehicle trajectory data shows the movement information of the object over time.Moreover,it can discover the traffic hotspot area through the gathered road segments or regions to discovery the urban traffic travel rules,which is help to provide traffic management with decision-making information.However,the spatio-temporal sequence characteristics of the trajectory make the traffic hotspot area change dynamically with time.The traditional clustering method is limited to the spatial attribute of the trajectory,and can not detect the distribution of the hotspot area with time,making it impossible to mine the travel law of moving object more accurately.Aiming at the above problems,this thesis proposes a road network hotspot area mining system based on spatio-temporal trajectory in mobile environment.The main research contents are as follows:(1)Trajectory data preprocessing model.It includes noise filtering,stop point detection,trajectory compression and segmentation.Firstly,based on the density clustering algorithm,the noise points are quickly filtered.Then,according to the temporal and spatial characteristics of the trajectory,the traditional DBSCAN clustering algorithm is improved to extract the stay points for reducing the useless data.Finally,through the method of curve boundary points detection,we extract the trajectory feature points as the trajectory compression and segmentation standard to reduce the storage pressure of the trajectory data and improve the data analysis ability.The above three steps help to provide effective data for the next trajectory clustering;(2)Spatio-temporal trajectory clustering algorithm ST-CFSFDP based on density peak.Considering that complication and uncertainty of vehicle trajectory distance calculation,this thesis redefines the trajectory similarity measure method,and draw on the ideas of clustering algorithm CFSFDP(clustering by Fast Search and Find of Density Peaks).A density peak trajectory clustering algorithm ST-CFSFDP based on spatio-temporal similarity metric is proposed,which improves the clustering center selection strategy by optimizing the domain search radius to ensure the clustering accuracy and reduce the artificial participation.And the final experimant proves that ST-CFSFDP has a better clustering effect than traditional spatial clustering algorithms.(3)Hotspot area mining system based on vehicle trajectory.Based on the above proposed algorithm and Beijing taxi GPS trajectory data,the system integrates data preprocessing,cluster analysis and hotspot extraction functions to explore the spatio-temporal distribution of hotspot areas,which reflect the urban traffic travel and provide important reference information for the formulation of relevant urban planning and traffic management measures.