Fast Production of Virtual Reality City Models

Virtual reality applications in the context of urban planning presume the acquisition of 3D urban models. Photo realism can only be achieved if the geometry of buildings is represented by a detailed and accurate CAD model and if artificial texture or real world imagery is additionally mapped to the faces and roofs of the buildings. In the approach presented in this paper height data provided by airborne laser scanning and existing ground plans of buildings are combined in order to enable an automatic data capture by the integration of these different types of information. Afterwards, virtual reality city models are generated by texture processing, i.e. by mapping of terrestrial images. Thus, the rapid acquisition of 3D urban GIS is feasible. 1 GENERATION OF VIRTUAL REALITY CITY MODELS The acquisition of three-dimensional databases for urban areas has become a topic of growing interest to the photogrammetric community. Simulations that require the representation, management and analysis of threedimensional descriptions have become standard applications for planning purposes in built-up areas. One example is the propagation of electro-magnetic waves that can be simulated for planning optimal locations for transmitter stations. A very widespread application is the computation of synthetic views or even animated fly-throughs to visualize urban scenes. The creation of a 3D city model for these virtual reality applications usually consists of a geometric object reconstruction followed by texture mapping to obtain a photo realistic model representation. Objects relevant for the 3D description of urban areas are buildings, streets, urban vegetation like trees and Digital Terrain Models (DTM). Because manual interpretation during data acquisition is very time consuming, a lot of effort has been spent to speed up this process by automatic or semiautomatic procedures. In our opinion an efficient acquisition of 3D urban models presumes the use of multiple data sources since a high degree of automation can only be reached by integrating different and complementary types of information. For the geometric data acquisition we use existing 2D GIS, color images and Digital Surface Models. As the information of a Digital Surface Model is restricted to surface geometry, the interpretation of this kind of data is easier e.g. compared to the interpretation of image data. A DSM, i.e. a geometric representation of the terrain surface, including objects like trees or buildings which rise from the ground, can e.g. be obtained from stereo image matching. For our purpose DSM from airborne laser scanners are utilized, since the direct height data acquisition by these systems provides DSM of high and homogeneous quality in urban areas, which is essential for our approach. Despite the high quality of the laser data, the automatic interpretation of DSMs remains a very pretentious problem. Within the approach presented in this paper the 3D reconstruction of the buildings is supported by given ground plans. This type of information is frequently available from 2D Geo Information Systems (GIS). By integrating the ground plans into the processing also the consistency between the already existing 2D GIS or map data and the generated 3D data can be guaranteed. One problem arising when existing databases are used is their potential lack of actuality and incompleteness. While aiming on the combination of a 2D GIS with directly captured data like DSM or images, a map revision has to be performed as a first step. For this purpose obsolete or incomplete parts of the GIS have to be uncovered. In addition to the detection or validation of inconsistencies between the datasets, there is a need to capture objects, which are not contained in the 2D GIS. While outlines of buildings and traffic network are available from standard databases, vegetation is usually not represented in detail. However, these objects are relevant for 3D site models. In order to automatically detect objects of interest like buildings, streets and trees in the DSM and multispectral images a classification approach is applied. To achieve photo realism an accurate geometric and textural description is required for each building. Since real images embody a representation of object detail they can substitute for geometric modeling. For this reason the use of photo realistic texture enhances the perceived detail even in the absence of a detailed geometric model. Hence, even though artificial texture can be assigned to building faces depending on the surface material, for realistic results real imagery has to be mapped to the facades at a sufficient resolution. In airborne data mainly the roofs of buildings can be observed. Due to the viewpoint, facades, jut out of roofs or passages are hardly visible or even totally occluded. The limited resolution of airborne data additionally restricts the amount of detail and the accuracy. Photo realism can only be achieved in the framework of virtual reality applications if terrestrial images are acquired and mapped on their corresponding building facades. After discussing the potential of airborne laser scanning for D. Fritsch, M. Englich & M. Sester, eds, 'IAPRS', Vol. 32/4, ISPRS Commission IV Symposium on GIS Between Visions and Applications, Stuttgart, Germany.