Graph-Based 3D Building Semantic Segmentation for Sustainability Analysis

A graph-based method is proposed to segment the 3D building models into semantically independent components. For each building, we first create a graph (N, E) in which the nodes N represent the surface of the 3D building model and the edges E standard for the shared lines between two surface nodes. Then, the graph is simplified by aggregating the connected coplanar surfaces. Next, the articulation points of the simplified graph are detected and removed literality to segment the graph into biconnected components. The semantic attributes of each component are detected according to its geometry features and spatial relationship with others. Finally, the building components with semantic and geometry information are used to simulate the city sustainability process such as energy consumption. According to the experimental results, the proposed method can effectively extract the semantic data from the LoD3/LoD2 building models for sustainability simulation tools such as EnergyPlus.