A 3D representation of a city embedded in an Internet environment is not a new technology. It is already used in many projects which present tourist information. But 3D modelling assumes a lot of manual operations and these are still not fully operational to handle, even if there exist special software (e.g. CyberCity) nowadays, that support the generation of 3D scenes. The emphasis of this paper is to present a method for improving the automation of generating 3D VRML scenes by using simple and adapted procedures which are implemented in a GIS environment. Our goal was to create a backdrop which is easily modifiable and replaceable. Especially within a city we find areas (e.g. shop windows), whose appearance changes continuously. Their representations can be replaced by simple work steps in the 3D workflow. A 3D environment is based on a pool of different data e.g. cadastral maps, floor plans, and aerial images, which should have a minimum resolution of 50 cm, to guarantee the recognition of all important items. The cadastral maps should be controlled by using current aerial images to eliminate errors. More complex buildings are assembled from several simplified building sections. Subsequently, an automated modelling takes place after adding a few parameters (building height, roof form, etc.).
For a realistic natural representation, the texturizing was assisted by using digital photographs of the building facade. The whole VRML scene is modular. Each wall was generated as an individual VRML file, turned and shifted into the suitable position. Every single wall can be addressed individually and this facilitates the adjustment of the textures enormously. In this way small changes of the buildings could be easily maintained without generating the whole VRML scene again.
Finally, the modular city file was loaded into a main file with optional additives such as trees, viewpoints etc. This additional information could also be generated likewise automatically.
For orientation purposes a general map was integrated, in which the position and line of sight of the user is indicated by an arrow.
In conclusion, it can be stated that the proposed method and its implementation have been proved to be very valuable and reliable for automated 3D urban modelling.