Procedural modeling of buildings

Creating 3D content remains a significant challenge in many application fields such as entertainment, architectural design, mapping, cultural heritage, or education. This applies especially for 3D models of urban environments, which are among the largest and most complex. And as 3D media moves mainstream and corresponding hardware capabilities tremendously improve, procedural modeling techniques are becoming an increasingly important supplement to traditional modeling tools. In this thesis, we propose the use of procedural methods for the design and modeling of buildings. Buildings are systems of high structural, spatial and functional complexity and their appearance is specified through arbitrary design choices made by architects. Modeling such complex systems has a long tradition in computer graphics, but unlike the simulation of natural phenomena where biologically motivated approaches are applied, the procedural reconstruction of human-made designs like buildings poses a number of novel problems. Besides structural and functional requirements, a system for the procedural modeling of buildings has to be able to encode a broad variety of architectural formalisms. The proposed procedural modeling technique is based on the concept of shape grammars. Shape grammars have been studied mainly in the context of architectural design theory, while their use in geometry generation is largely unexplored. To close this gap, we introduce the CGA shape grammar, a shape grammar which is amenable to computer interpretation and suited for the procedural generation of detailed 3D building geometries. The CGA shape grammar is a symbolic shape grammar that enables the definition of rules that iteratively evolve a design by creating more and more details. In the context of buildings, the rules first create a crude volumetric model of a building, then continue to structure the facade and finally add details for windows, doors and ornaments. We present the complete definition of the CGA shape grammar including its concise notation, the definition of the shape operations, and examples detailing various modeling strategies. An important task is to reverse engineer the CGA shape grammar rules from existing buildings. Therefore we present an automatic method to analyze and reconstruct single facades. We mimic the procedural modeling pipeline to subdivide a facade in a top-down manner into elements such as floors, tiles, windows, and doors by using image analysis methods. This results in a CGA shape grammar representation. The above methods are implemented in the CityEngine, a unified framework for the authoring, generation and editing of procedural urban environments. Based on the CGA shape grammar, this system defines a complete and versatile modeling environment for the creation of 3D buildings or cities.