Procedural city model using multi-source parameter estimation

Most current digital 3D city modelling procedures have either a low degree of automation or require specialized skills. Moreover, the construction process is the result of an equilibrium between the desired level of detail on the one hand and modelling performance on the other hand. Although environmental 3D models and 3D city models in particular are essential for a wide range of applications and disciplines, these difficulties are substantial bottle necks for the availability of the models. In this paper, initial steps and ideas behind a novel approach for the construction of 3D city models are presented using an Airborne Laser Scanning (ALS) point cloud and standard digital 2D data. The first step involves point processing and feature detection for an ALS point cloud, resulting in the separation of building and ground points from vegetation and other points in the point cloud. Secondly, the detected building features are described in more detail using the 2D data, allowing the distinction between roof points and façade points. A texture map is assigned to the detected features using image libraries. The 2D data are also used for the improvement of vegetation mapping. The novelty of this approach is the fact that the actual city modelling is performed using recently made available software. The used software allows the interpretation of conceptual rules for the automated modelling of real-world environments. The proposed workflow is illustrated by the construction of a city model of some part of Geraardsbergen (Belgium).