Dynamic construction of visible knowledge graph based on 3D geographic scene for intelligent bridge management

Abstract. Knowledge graphs are a crucial core in enhancing the intelligent management of geographic entities. Current knowledge graphs primarily focus on constructing knowledge in bridge health monitoring, often neglecting the spatial relationships of bridges as geographic entities. This oversight limits their capabilities in diagnosing, analyzing, and predicting bridge conditions. To address this, a method for dynamic construction of knowledge graphs is proposed, which is driven by semantics and data for extracting entities and relationships from geographic scene. Furthermore, a multi-level 3D knowledge graph visualization method was proposed, presenting entities and their relationships in varying levels of detail. Experiments and analyses were conducted using a geographic scene that included a building information model of a bridge. The results demonstrate that our proposed method can construct knowledge graphs both rapidly and accurately, providing a more comprehensive understanding of the relationships between bridge entities. This enhances the management of intelligent bridges.