The Research Of Spatial Database Oriented To Virtual City Application

Nowadays with the tremendous advances in computer graphic, physical storage capacity and computer network, Virtual City Application, which bases on these technologies, has become possible. The basic problems in realizing Virtual City are how to store, load, query and update large-scale 3D models efficiently, and how to manage 3D Model with the vector-graph, text attribute and raster data. Traditionally, the 3D Model is store as file. But with the data size become huge, there are some draw back with this storage scheme, such as loading slowly, hard to update and hard to query sub-scene. This thesis presents the design and implementation of a DBMS-based 2D/3D data storage scheme, compatible with VRML/X3D data model.Currently, DBMS bases on Relation Data Model, which is not very suitable for spatial data. In our implementation, we use the Oracle Object-Relation extension and its spatial data component Oracle Spatial as the basic platform. We use the data type SDO_GEOMETRY to deal with 2D vector-graph. According to the VRML/X3D specification, we design a set of tables to store the scene-graph's nodes and the relation between the nodes. We also develop a set of tools to import data from WRL files, 3Dmax and AutoCAD. In our test, loading data from database is 4~5 times faster as loading data from WRL files when the data size larger then 1Gega-byte.One of the most important features of database system is it can support indexing and effective query. Most databases Using B+ tree for number and text. For spatial indexing, we use quad-tree and R tree, which are supported by Oracle Spatial. To create spatial indexing, firstly iterate the scene-graph of 3D Model and calculate the bounding-box of each shape;then we project the bounding box to the X-Z plane and get the 2D contour of each shape and store as SDO_GEOMETRY;the last step is create spatial index using SDO_GEOMETRY field.