Research And Application Of Trajectory Clustering Method Based On Spatio-temporal Constraint

With the development of wireless communication, internet, global positioning, measuring and sensor technology, research on information processing and knowledge discovery from moving objects have attract people's attention, which can be applied in many area like intelligent transportation system, mobile service system and environment monitoring system. The explosive mobile data, which could be acquired at anytime and anywhere, bring us a lot of information; however, the extrusive problem"massive data makes a hard knowledge discovery process"become more and more serious. Recently, more and more researchers realize that most of the movements of moving objects are spatio-temporal constraint; research on the trajectories of moving objects brings more knowledge than working on the discrete location data. For example, we can perform motion analysis and prediction through trajectory mining to support navigation service.The purpose of this study is to find the routine and behavior of the motion of moving objects by analyzing spatio-temporal feature of trajectories and applying clustering method to find similarity trajectories. With the knowledge discovered from trajectories, we can provide more useful information for navigation, route selection and monitoring.This thesis describes the current research status towards trajectory data mining, and proposes a new trajectory clustering method based on trajectory partition and spatio-temporal similarity measure by analyzing spatio-temporal features of trajectory data and the weak points of the existing methods. We first present a spatio-temporal semantic enriched trajectory and perform the characteristic points choosing and partitioning on it. Then, we propose to measure the spatio-temporal similarity and distance between trajectories. Finally, we improve the sub-trajectory clustering algorithm to discover knowledge behind trajectories. The analysis and experimental showed that the proposed method can effectively reduce the memory use of trajectory processing. It also exhibited its efficiency in spatio-temporal similar trajectory searching and scalability in discovering spatio-temporal closing trajectories. On the basis of the theory, this thesis finally design and implement a trajectory analysis prototype of intelligent transportation system based on trajectory clustering method. It methods can effectively analysis spatio-temporal constrained trajectories and visualize the clustering results.