Research Of Partition And Clustering For Trajectories Of Moving Objects

With the development of satellites, tracking facilities and telecommunication technologies the tracking ability for continuous moving objects is enhancing. It is possible to collect a large amount of trajectory data of moving objects. Since the trajectory of a moving object contains a lot of information, it is an interesting task to analyze trajectories for several application areas. Examples include vehicle management, weather forecast, intelligent navigation, biologic medicine, commercial decision, military analyses and antiterrorist monitor. It is one of typical data analysis task to cluster similar trajectories of moving objects and extract characteristic movement model in order to analyze and forecast behaviors of objects.A trajectory of moving objects may have a long and complicated path, there may be similar in small portion but be dissimilar as a whole. For example in traffic monitoring many people will pass through the same main sections, however their final destinations are not same. Distance measure is a major factor to affect results of clustering, the traditional clustering method regard the trajectory of this particular track as a whole, the distance between the trajectories contains information of all the trajectory points. The distance between the trajectories with similar paragraph may be different, can not be gathered into a cluster, and this similar paragraph is missing. The purpose of our research is to find out these similar paragraph called as sub-trajectory.Proposed framework is to partition a long trajectory into a group of straight line segments, and then cluster the similar line segments, finally extract common sub-trajectory from the straight line segment clusters. The contributions are as follows:1. In order to deal with trajectories with different speed at different temporal we used folding line segment to represent data model of trajectory. Trajectory moves along straight line until it changes direction or speed. Add in time and speed dimensions to the space dimension.2. Using hierarchy distance to improve clusters of partitioned line segments. Filtering similar line segments based on spatial distance similarity and refining clusters based on temporal and speed distance. 3. Proposing an algorithm to partition a trajectory into a group of line segments. We consider not only changes of direction but also changes of speed.4. Clustering the trajectories partitioned line segments based on OPTICS algorithm and finding out clusters of similar line segments.5. Discussing theoretical analysis and related algorithms. Analyzing the experiments and verifying in spatial-temporal datasets. Experimental results demonstrate that the framework can correctly discover common sub-trajectories from real trajectory data and are efficient and scalable.