In autonomous driving, traffic sign detection is easily affected by environmental lighting, changes in target size, and complex backgrounds. To address these issues, this paper proposes a multi-scale traffic sign detection algorithm, SS-YOLO (Small Sign). This method enhances multi-scale feature fusion by generating additional high-resolution feature layers and optimizes small object detection by removing the large-object detection head. At the same time, the C3 module is replaced with the more efficient C3-DS module, and the EMA attention mechanism is introduced to improve the ability to capture channel information. Additionally, the DySample upsampling operator is used in the feature fusion layer to increase the receptive field and optimize feature fusion. Experimental results show that SS-YOLO achieves an average precision (mAP) of 86.8% on the CCTSDB 2021 dataset, a 10.5% improvement over YOLOv5s, while reducing the number of parameters by 23.23%, achieving both fewer parameters and higher detection accuracy.