1. Introduction
Our human senses evolve and are geared toward perceiving the world in three dimensions (3D). In today’s marketplace, sophisticated 2D GIS are available as a standard and it seems more practicable, wherever possible, to extend the 2D system to handle 3D applications rather than building a brand new 3D system from the scratch. Luckily, when considering the data representation in geometrical terms, 2D is a subset of 3D and therefore in principle, it should be possible to extend GIS from 2D to 3D and such an extension brings several challenges, particularly in constructing the 3D data model, storing data, indexing and finally the provision of an interface to visualize data. In this paper the focus is on the issues of storing and indexing 3D data as well as object query. The system that we have used to extend is the database structure of Intergraph GeoMedia (Figure 1). The paper starts with an overview of storing 3D data in the database that was initially designed to store data of 2D. The 3D data model is then defined by using Delaunay Tetrahedral Tessellation (DTT)(Chen and Murai, 1999). This is followed by a discussion of data compression, as even moderately fine tetrahedral meshes consume large storage space. The next section discusses indexing method, particularly the structure of Hybrid Tree (Chakrabarti and Mehrotra, 1999). In order to get the Hybrid Tree structure to work with the DTT model, the respective data is mapped as points into high dimensional feature space. Following the explanation of the indexing method, the algorithms for data searching inside the index tree is then explained. Finally, the results of a trial run are reported as well as suggestions for future work

2. Related Works
The topic of extending GIS to the third dimension (3D GIS) received much attention and of the literature is now extensive. However, most of the proposed systems are characterized by dealing with some specific aspect of 3D and have yet to offer a holistic approach that deals with all aspects of 3D GIS. For example, special attention has been given to 3D visualization that linked to the conventional GIS and this has resulted in the availability of commercial products such as “SiteBuilder3D” (SiteBuilder3D, 2003). This is one of the products developed by Multigen-Paradigm (Multigen-Paradigm, 2003) that provides high-end 3D visualization based on a GIS database, but does not provide analysis capabilities. Other products, such as 3D Analyst of ESRI’s ArcGIS (ESRI, 2003), do provide analysis functionality, but its visualization capabilities are limited. Yet another product called GeoMedia Terrain developed by Intergraph (Intergraph, 2003) serves as Digital Terrain Model (DTM) module for the GeoMedia GIS on terrain model generation, terrain analysis and fly-through. Nevertheless it is based on a 2D datasets and should be considered as 2.5D instead of 3D solution. Apart from the traditional GIS vendors, some other companies have recently provided GIS module in their product. For example the Imagine system, originally developed by ERDAS (ERDAS, 2003) for remote sensing applications has now a GIS module (Imagine VirtualGIS). This module enables 3D analysis to be carried out together with visualization; nonetheless, the analysis is mainly based on visual instead of object.


Figure 1: Database structure of GeoMedia for Microsoft Access. There are five columns in total. [ID] is the index column for the entire row with “Number” type, [RIVER_NAME] and [RIVER_TYPE] are the attribute columns with “Text” type, [Geometry] is spatial column with “OLE Object” type, and [Geometry_sk] is index column for spatial data with “Text” type
Whilst these developments are mainly coming from the GIT community, there are also some researches in computer science or engineering that are working on spatial data storage issues and their work is highly relevant to 3D GIS. In the following paragraphs, we review relevant work that deal with data model, storage and indexing.

Page 1 of 5
| Next |