To address the issues of severe fragmentation of Traditional Chinese Medicine (TCM) prescription knowledge and single knowledge representation, a standardized and visualized prescription knowledge graph is created by extracting the medical case data and prescription information based on the TCM medical cases and prescription codices, and incorporating external knowledge, such as prescription ion-evidence relationships. The collected and organized prescription data is initially recognized and extracted from the entities to obtain attributes like prescription name, composition, function and treatment, symptoms, etc. Second, a knowledge graph of prescriptions is constructed top-down, and Neo4j's Cypher is used to manipulate and visualize the prescription data. Finally, the tongue image samples are trained for tongue image discrimination using the Faster R-CNN of deep learning. Based on the fusion of tongue results, a prescription knowledge graph recommendation system for tongue diagnosis is created. To improve the clinical analysis of TCM diagnosis, the system can use to carry out operations like querying prescription information and automatically generating prescription recommendations based on the results of the tongue diagnosis. The results in the realization of personalized treatment method ds in TCM and a new idea and method for prescription recommendation in TCM.