Abstract: At present, the research on the integration of BIM and GIS is mainly divided into three directions, namely data format conversion, data standard extension, and ontology. This project chooses to carry out the entity integration of IFC and CityGML through the ontology mode. The triples of entities are extracted through semantic modification, and then the triples are represented as entity vectors by knowledge representation learning. The semantic information of entities is transformed into a low-dimensional vector space, and the matching fusion system of IFC and CityGML standards is constructed based on extended semantic matching. To solve the problem of semantic understanding bias of the same entity under IFC and CityGML, and to cluster by similarity calculation. In addition, the extracted triples were used to carry out geometric reconstruction using entity-relations-entity, and the entities of IFC standard were mapped to entities of CityGML standard. Finally, the established 3D model is used to verify and to realize the transformation from BIM field to GIS field.