Geological maps and their ancillary resources (e. g. , related texts and tables), as an important storage medium for geological knowledge, contain a large amount of geoscientific knowledge as well as expert empirical and implicit knowledge. Rapid and accurate knowledge specification definition, formal representation, semantic association and precise reasoning for multimodal geological data are the frontiers of current geoscientific knowledge extraction and mining. Currently, the knowledge distribution of maps and ancillary descriptive information in massive geological data is scattered, and the information in the maps cannot be effectively associated with the descriptive information and provide knowledge services. Methods: In this paper, we take multimodal geological data as data source, establish a geological map knowledge expression model driven by multi-source data, express explicit and implicit geological knowledge embedded in maps through geological knowledge, rule knowledge and decision knowledge, and express relationships among geological objects from three dimensions of semantic, spatial, and attribute; we also carry out a study of the precise extraction and storage of geological entities and relationships based on domain knowledge for geological maps and descriptive information, and finally conduct a study on the precise extraction and storage of geological entities and relationships based on the domain knowledge for geological map and descriptive information. Result: The vector data of 1∶50, 000 Mineral Geological Survey of Yinkeng, Yudu County, Jiangxi Province on map sheet G50E011007 is taken as an example for validation and analysis. The experimental results verify that the knowledge expression model and knowledge extraction method proposed in this paper can better solve the problems of weak graphical association and insufficient knowledge mining in geological knowledge modeling.