The visual identification of intangible cultural heritage (ICH) inheritors enables rapid retrieval of key information related to intangible cultural heritage and promotes its digital preservation and dissemination. Currently, most provincial and municipal ICH dissemination platforms present inheritor information primarily in list form. Visual business card data remain relatively limited, and accurate recognition of visual information requires the joint consideration of semantic features and spatial layout. In this study, a visual information recognition method named Semantic-Graph, which integrates semantic extraction with graph feature enhancement, is proposed. First, by integrating information collected from public websites, a visual business card dataset of ICH inheritors in Shaanxi Province was constructed, covering ten categories, including ICH names, regions, and descriptive content. Second, an ICH-specific lexicon and a semantic extraction method were constructed to obtain textual information from visual business cards. Finally, a graph feature enhancement strategy was introduced. The robustness of the model was improved through random node masking and random edge deletion, while the positional relationships among nodes were learned through a positional attention mechanism to enhance the model’s recognition capability in visual space. Comparative experiments show that the recognition performance of Semantic-Graph is significantly higher than that of other benchmark models, achieving a macro-average F1 score of 0.9281. Ablation and parameter influence experiments further verify that the proposed graph feature enhancement method effectively improves model performance, with an 11.6% increase in F1 score compared with the baseline model. In addition, Semantic-Graph achieves superior performance in comparative experiments conducted on multiple publicly available datasets.