The overall project aims at the creation of a 3D real-time visualization system for large-scale Geographic
Information Systems (GIS). It is the goal of this project to interactively navigate in virtual environments that depict
detailed landscapes of selected cities or rural areas. A user can control a viewpoint to fly over or walk through the
virtual environment.
Current large-scale systems mainly use aerial or satellite images mapped on height field data to model the
environment. However, newer systems include many objects, like trees, bushes, and buildings, in the interactive
simulation. If many of these objects are placed on the terrain the size of the database explodes and new problems
arise.
The problem connected with the fast rendering of large-scale databases is that the huge amounts of data defeat any
nave approach of rendering. A crucial part of fast rendering is to have several versions (levels-of-detail) of objects
or scene parts. If an object is important (e.g. because it is close to the viewpoint) it will be rendered in high detail
but less important objects (i.e. because they are farther from the viewpoint) can be rendered with less detail. We
propose to develop new simplification methods especially suited for 3D GIS systems. The proposed solution falls
under the category of image-based simplification. We want to answer the following questions:
How can numerous objects that populate the surface of the earth, like trees, bushes, buildings and street features be
represented by simple image-based representations that allow fast rendering?