Declustering of Trajectories for Indexing of Moving Objects Databases

Efficient storage and retrieval of trajectory indexes has become an essential requirement for moving objects databases. The existing 3DR-tree is known to be an effective trajectory index structure for processing trajectory and time slice queries. Efficient processing of trajectory queries requires parallel processing based on indexes and parallel access methods for the trajectory index. Several heuristic methods have been developed to decluster R-tree nodes of spatial data over multiple disks to obtain high performance for disk accesses. However, trajectory data is different from two-dimensional spatial data because of peculiarities of the temporal dimension and the connectivity of the trajectory. In this paper, we propose a declustering policy based on spatio-temporal trajectory proximity. Extensive experiments show that our STP scheme is better than other declustering schemes by about 20%.