[Objective] In order to inherit and promote the seal culture and improve the recognition effect of seals in complex situations, the recognition results and related knowledge are structurally displayed in combination with the knowledge graph and visualization technology. [Methods] A deep learning model integrating multiple features is proposed. The model extracts the color feature graph, edge feature graph, and gray level feature graph of the seal image and then inputs the three feature graphs into the deep learning model for recognition. Finally, the identification results are compared with the nodes in the knowledge graph, and the relevant knowledge is displayed. [Results] This paper collects and tags the seals contained in thirteen calligraphy and paintings, such as“Cold Food Posts”, and takes two of them as test-sets. Compared with VGG16, P, R, and F1 increased by 28.40%, 28.67%, 28.54% respectively. Without fusing multiple features, P, R, F1 decreased by 24.30%, 20.16%, 22.74%. [Limitations] The model can only extract and recognize the global features of the seal, and lacks the ability to identify and reason the local semantic information of the seal. [Conclusions] The experimental results show that the proposed method has a good effect on seal recognition tasks, and the multi-dimensional feature graph can improve the recognition ability and robustness of the model to complex situations.