ON THE USE OF GEOMETRIC AND SEMANTIC MODELS FOR COMPONENT-BASED BUILDING RECONSTRUCTION

Abstract

3D building data is needed in many application areas. Besides the geometric description an increasing number of applications also demand thematic information about acquired buildings. We present a concept for the automatic extraction of buildings from aerial images. In contrast to other approaches generic building structures both are geometrically reconstructed and semantically classified. A component-based, parameterized building model is employed to control the reconstruction of buildings. This paper describes how geometric and semantic knowledge of buildings is propagated through the different aggregation levels of the building model. Furthermore, it is shown how rules and constraints are derived from the model and exploited at each stage of the reconstruction process. 1 INTRODUCTION 3D building data play an important role for digital city models and geoinformation systems, which are used in numerous application areas: architecture, urban planning, telecommunication and environme...