Multi-resolution representation of 3D complex building models with features preservation

Automated constructing multi-resolution 3D building models is an important issue in the contexts of virtual reality, progressive transmission and 3D GIS. This paper proposes a novel method to generate multi-resolution representations of the 3D complex building model with features preservation based on cell clustering. The quadratic error function is constructed to efficiently measure the clustering error of each cell. First, the 3D complex building model is converted into volumetric representation. Then, the boundary cells are progressively clustered into a coarser version and a corresponding cell binary tree is generated. Finally, the simplified building model can be extracted according to a modified dual contouring algorithm. The experimental results show that the proposed approach provides an efficient and robust solution for the automated generation of multi-resolution building models, with capability of preserving the dominated structure features.