Aiming at the problem that the accuracy and real-time performance of traffic sign recognition in natural scenes can’t meet the actual requirements, an improved YOLOv3 traffic sign recognition algorithm is proposed. Due to the low accuracy of predicted target frame in YOLOv3, attention mechanism is introduced to improve the border regression. The edge features of the target object are weighted heavily in order to improve the accuracy of the predicted target frame. Finally, the improved YOLOv3 algorithm is applied to traffic sign recognition. According to a large number of experiments and comparative analysis, the YOLOv3 algorithm combined with attention mechanism not only ensures the accuracy of recognition, but also improves the accuracy of the predicted target frame. When applied to traffic sign recognition, it achieves better recognition performance.

CICTP