| With the development of science,technology and economy,various sensing devices have been widely applied.With the enhancement of people’s living levels,the smart phones,smart watches and other smart devices are increasingly popular.People typically travel with some smart devices,and their travel manners become diverse,such as walking,riding and driving,taking buses and subways in cities,and taking trains and planes for long-distance travels.The carried smart devices can collect a large amount of trajectory data.Through mining these trajectory data,much valuable information can be obtained for many applications,such as activity recommendation,urban planning,public security,national military defense and so on.However,with the exponential growth of trajectory data,the trajectory mining faces many new challenges.First of all,massive data produces a large storage overhead,and its transmission produces a long transmission delay.The network bandwidth is largely occupied as well.It is very hard to store the exponentially growing size of data.In addition,trajectory mining queries a large number of computations.Typically,the massive trajectory data contains much redundant data,so that through trajectory compression or trajectory partition,the redundant data can be found and removed,thereby reducing the required communications,storage space and computations.The existing trajectory partition methods try to find and remove the redundant trajectory points,and reserve the valuable trajectory points.Different movement features can reflect the trend and law of node movements from different views.However,the existing trajectory partition methods methods do not partition the trajectories from the perspective of movement features,and thus the partition results are not preferable.In this thesis,the movement behaviors of nodes are analyzed in terms of velocity and accelerated velocity,and Trajectory Partition Method Based on Double Velocities(TPDV)is proposed.This method can effectively solve the problem that the original trajectory shape changes dramatically at the selected stop point.(1)In order to solve the problem of choosing wrong starting track points and failing to determine stopping points from the tracks with round-trip paths,in this thesis,the acceleration is introduced into SPEMS,and Stop Points Extraction Method based on Double Velocities(SPEDV)is presented.(2)In TPDV,the changes in the moving velocity of nodes are first detected to find the feature points,and according to the changes in accelerated velocity the accelerated velocity-changing points can be extracted.Then,the stay points can be determined according to the time and scope of node activities,and the trajectories are partitioned by the obtained feature points,accelerated velocity-changing points and stay points.Extensive simulations are carried out based on the Geolife trajectory dataset,and the simulation results show that the proposed trajectory partition method based on double velocity features performs well in terms of running time,simplification rate,and partition error under different loads.(3)Finally,theoretical design of the TPDV algorithm and the effective verification of the experimental results,this thesis also designs and implements a prototype: vehicle trajectory partition system,which can implement TPDV.In chapter 4,the overall design and implementation process of the prototype system are discussed in detail.The running results of the prototype system show that this system can effectively divide the trajectory data and visually display the results of trajectory partition on the online map. |
