Generalization and visualization of 3D building models in CityGML

Abstract

Generally, cities are expanding due to rapid population growth and require 3D city models for effective town planning, communication and disaster management. Rendering of 3D scenes directly is not so much appropriate as appearance properties, textures and materials attached with city models drastically increase the loading time for visualization and spatial analysis. Additionally, different applications or users demand different Level of Detail (LoDs), thus one of the questions arises - how different LoDs can be made available to these applications? Generation of lower LoDs given by OGC standard CityGML from higher LoDs to reduce data volume is a generalization problem. Relying only on existing geometric-based generalization approaches can result in the elimination or merging of important features, hence, semantic information can be considered. A review of pertinent generalization algorithms proposed by several researchers is presented. Additionally, this paper provides a method for generalization of 3D structures with the aim to derive multiple LoDs keeping semantic information into account. For this purpose, height and positional accuracy of objects at different LoDs provided by CityGML are considered. Initially, building parts and installations are removed. 2D footprints of remaining 3D structures are projected onto ground and simplified to derive LoDs building geometry. An adoption of methods of Sester and Brenner (Continuous generalisation for visualisation on small mobile devices. Heidelberg, pp. 355-368, 2004) extended by Fan et al. (Lecture notes in geoinformation and cartography, advances in giscience. Springer, Heidelberg, pp. 387-405, 2009) are applied for simplification and aggregation of projected footprints. The experiments showed that due to repetition of coordinates of connected nodes in CityGML increase both the rendering time and memory space. However, elimination of important smaller features can be avoided by taking semantic information into account while performing generalization operations. © 2013 Springer-Verlag Berlin Heidelberg.