With the rapid advancement of urbanization, the field of traffic signs is encountering severe challenges, including worsening traffic congestion, frequent accidents, and inefficient management. In the area of traffic sign detection, traditional algorithms demonstrate significant limitations in complex scenarios. Particularly under conditions such as rain, fog, low-light environments, or partial occlusion of traffic signs, their detection accuracy declines sharply, resulting in a large number of traffic signs failing to be identified promptly and accurately. This not only impacts the real-time acquisition of traffic information but also poses a serious threat to traffic safety. To address these issues, this study develops a multi-scenario traffic sign detection datasets, MSTS-DataSets (multi-scenario traffic sign DataSets), to cover traffic conditions across various scenarios. Additionally, based on the YOLOv11 network architecture and integrating dynamic convolution and the Area Attention mechanism, we propose the DCA-YOLO (Dynamic Convolution Attention-YOLO) network model. Compared to YOLOv11, the new architecture achieves a 6% improvement in mAP50 performance, with the final model reaching 93.2% mAP50.