A Study Of Indexing Techniques In Spatio-Temporal Database

In recent years, location based mobile e-services which provide quick and convenient information, are chased by service providers and users. As a result, software technologies that enable the management of the positions of moving objects are in increasingly high demand. It is a new research branch in database field, i.e. Spatio-Temporal Database.In traditional databases, data is assumed to remain constant unless it is explicitly modified. Capturing spatio-temporal information would entail either performing very frequent updates, which results in using too much system resource; or frequency of updating too small to get required accuracy. In addition, there are spatial relationships among the positions of objects, which can't be identified by traditional indexing techniques, such as distance, direction and range. So there has been great interesting in finding new indexing techniques for Spatio-Temporal databases.After surveying the existing access methods in spatial database and temporal databases, we classify the index models in spatio-temporal database into three type according to the time interval of the data that it manage: indexing of the current and the future position; indexing of the past position or trajectories, up until the most recent position sample; indexing of the past, current and the (near) future position. We discuss some index structures of these three types. After analyze the demand of market, a new indexing technique which is called Indexing Moving objects Trajectories on Fixed Networks (IMTFN) are proposed in this paper. IMTFN consists of a 2-dimensional (2D) R*-Tree for managing the fixed networks, a forest of 1-dimensional (1D) R*-Trees indexing the time interval for managing the position of moving objects, and a hash structure for the newest location of moving objects. IMTFN supports the efficient query of the present and past positions of moving objects, optimizes operations of windows query and trajectory query. Extensive experiments are conducted to evaluate the performance of the proposed indexing mechanism and show that IMTFN perform considerably better than STR-Tree and FNR-Tree.