Research on Lightweighting Methods for 3D Building Models Based on Semantic Constraints

Abstract. Three-dimensional building models, as core data in Building Information Modeling, are extensively applied in various fields such as architectural design and urban planning. However, the geometric complexity and the vast amount of data involved pose significant challenges for storage, computation, and efficient application. Therefore, ensuring the geometric accuracy and semantic integrity of models while reducing data redundancy has become a critical issue in current research. To address this problem, this paper proposes a lightweighting method for 3D building models based on semantic constraints. The method combines face merging and face movement graphical simplification algorithms to simplify models from LOD3 to LOD1 in a layer-by-layer manner. In the transition from LOD3 to LOD2, a face merging algorithm is employed to analyze the semantic consistency of adjacent faces, ensuring that only semantically consistent faces are merged, thus generating an LOD2 model with semantic integrity. In the transition from LOD2 to LOD1, a concave-convex face movement and merging algorithm is utilized to reduce redundant data while maintaining geometric similarity and optimizing the model's topological structure, ultimately producing a lightweight LOD1 model. Experimental results demonstrate that the proposed method significantly reduces the model's storage requirements while effectively preserving its semantic information.