Research And Implementation Of Spatiotemporal Database Indexing Technology

With the development of the related technologies such as mobile computing, Global Positioning System, GIS, etc., database need store and manage large quantities of physical objects with spatiotemporal information in realistic world, and their spatial location or scope will vary with time, which will promote the development of spatiotemporal database. The application of spatiotemporal database spreads over many domains as transportation (such as vehicle supervision), weather monitoring, military and so on. In the spatiotemporal database system, indexing mechanism is a key technology to ensure the effective deposit and withdrawal of the spatiotemporal objects and thus has become the research focus of spatiotemporal database.Currently there is not a highly efficient indexing technology which is universally applicable to all environmental requirements. The 3D R-tree, based on R-tree, regards time as another dimension. Its searching process is very direct-viewing and avoids the differentiation between time search and space search, more suitable to the spatiotemporal objects whose indicating position and scope do not vary or vary little with time. However, it fails to consider the special nature of the time dimension, and can only handle the off-line data. Moreover, for those long-term static objects, many long cubes will be formed, making the indexing function descend dramatically. For this, this paper proposes the idea of giving the priority to the consideration of the fission method along with the time axis, to reduce the number of the long cubes; to realize the "on-line" data index through separating the indexing of the historical data and the active data, finally forming 3D R*-tree~ the expanded version of the3D R-tree, and proposeing a new cost model for optimizing 3D R*-tree, with its timing function enhanced by 40%, its time section inquiring enhanced by 30% compared with HR-tree. Though the time section inquiring is slightly low, the use of space has dropped by 40%, achieving the best integration of 3D R*-tree and database, carrying out the indices of spatiotemporal database successfully.The main contributions of this dissertation are as follows : 1) It proposes a "splitting mechanism". For the spatiotemporal objects with a long-time historical evolution, split them first along the time axis, considerably reducing the density of time object data sets and improving the indexing efficiency.2) According to the characteristics of online data, establishes two 3D R-tree: Active 3D R-tree and historical 3D R-tree, which enables the indexing mechanism to index active data.3) Combining the "splitting mechanism" with double 3D R-tree, proposes 3D R*-tree: it can index offline data and online data, and indexing performance improves remarkably.4) It proposes a new cost model for optimizing 3D R*-tree.5) Using expansion module of Informix the database server -Datablade module, it has realized the integration of 3D R*-tree and the Informix object database.