Data Analysis And Research Of Vehicle Trajectories

A spatial trajectory is a sequence of sampled points including coordinate and timestamp information which come from mobile objects. Recent years have witnessed the accumulation of tremendous trajectory data in daily life due to the developments of various locating systems like Global Positioning System(GPS) and WIFI,along with pervasive use of mobile devices. Mining trajectory data has been gaining significant interest.We focus on car trajetories on road networks and aim at trajectory clustering with experiments fulfilled on real data.Experiment results with certain value is acquired.Specifically,the main contributions and innovations are summarized as follows:(1)We optimize the SLAMM(Selective Look-Ahead Map Matching) algorithm which is very effective in processing trajectories from T-Drive project and get faster running speed with accepted accuracy.We adopt trajectories with high sampling rate from cabs in Beijing City.SLAMM is an algorithm good at processing high sampling rate trajectory with high speed and accuracy.We optimize the SLAMM algorithm by studing the constraints of road network and taking advantage of features in road network.(2)The effect on trajectory clustering imposed by constraints is studied,then a simplified NEAT model with time factor added is applied to process the trajecories to gain clusters on the road network,which implies the traffic situation and other information of the road network and finds itself quite useful in fields like transportation administration, personal route planning and so on.(3)We fulfill experiments with trajectory data from T-drive project on road network of Beijing City and visualize the resulting clusters of which we discuss the meanings.