The aim of this paper is to propose a new road network pattern recognition method, namely TransRoadnet.Unlike existing methods, the proposed method offers two key advantages. First, it is able to recognize multimodal hybrid road network patterns, which significantly improves the generalization ability of the model. Second, by integrating the Transformer model, it enhances the extraction of relational features between the overall structure of the road network and its road segments, thereby improving the accuracy of pattern recognition.