A Quantitative Texture Feature Preserving Method for Simplifying 3D Building Models

Three-dimensional (3D) city models have become an important component of spatial information infrastructure, and widely used in studies of virtual geographical environments, urban planning, and navigation, as well as other fields. Level of detail (LOD) representation of 3D models provides a foundation for the large-scale visualization and application of building models, and simplification is one of the most popular approaches used to generate multi-resolution 3D building models. This paper proposes a texture feature preserving approach to the quantitative simplification of manually produced building models. First, the original models are converted to polygonal models with the B-Rep method. The structural features of the buildings are then extracted and used to perform quantitative simplification of the building models while preserving the structural features. Next, the textures of new triangles are synthesized from the original textures. Finally, a post-processing algorithm is used to detect overlapping and/or intersecting triangles to improve the simplification results. Experimental results show that the proposed method can be adapted to models of ordinary buildings and landmarks. It can be directly applied to large-scale building model network visualization.