In the scientific research of crop breeding, traits are mainly controlled by key genes, and the cultivation of new crop varieties with the aggregation of a variety of elite traits has always been the direction of breeders. Therefore, the discovery of crop gene knowledge, that is, the knowledge related to crop genes, can provide new research ideas for the analysis of the molecular regulation mechanism of crop traits, and will effectively help the cultivation of new crop varieties. In the context of the scientific research paradigm of AI for Science, artificial intelligence technology has become the engine to help the innovation and development of crop breeding scientific research. As a knowledge organization technology in the field of artificial intelligence, knowledge graph technology can capture the complex correlation between genes and traits across species, and trigger the discovery of knowledge related to crop genes. Based on the in-depth analysis of the demand for crop gene knowledge discovery services, combined with the correlation characteristics of multi-dimensional crop breeding scientific data, selected the PubMed scientific literature database and eight field scientific databases as data acquisition sources, and uses multi-path knowledge extraction to construct a knowledge graph of cross-species and multi-dimensional scientific data fusion. Based on the knowledge graph, a crop gene knowledge discovery model was constructed, and the discovery of elite genes and gene function knowledge was realized. Finally, from the perspective of building a knowledge base supporting AI4S, building a technical system for intelligent computing, and building a knowledge service model driven by “data + AI”, the empowerment of artificial intelligence technology in agricultural scientific research is prospected.