Visual Data Mining in Large-Scale 3D City Models

This paper presents an approach towards visual data mining in large-scale virtual 3D city models. The increasing availability of massive thematic data related to urban areas such as socio-demographic data, traffic data, or real-estate data, raises the question how to get insight and how to effectively visualize contained information. In our approach, we extend a real-time 3D city model system by features that interactively map thematic data to specified graphics variables of the city model’s geometry and appearance. In particular, a rendering technique is explained that can efficiently represent and reconfigure the scene graph of a 3D city model even for large-scale models. The resulting dynamic 3D city models serve as general geovisualization tools to effectively analyze, explore, and present geometric and related thematic information of urban areas. Sample applications include city information systems, urban planning and management systems, and navigation systems.