Research on Road Traffic Sign Detection Based on Edge Intelligence

With the development of intelligent transportation and automatic driving technology, traffic sign detection is of great significance to ensuring driving safety and realizing intelligent perception. Facing the realistic demand of edge devices with limited arithmetic power, this paper proposes a lightweight traffic sign detection model based on improved YOLOv5, aiming to improve detection accuracy and real-time performance to meet the deployment requirements of edge computing scenarios. The kmeans++ algorithm is used to perform anchor frame clustering on a subset of the TT100K dataset to optimize the Anchor matching effect. At the same time, MobileNetV3 is introduced to replace the original CSPDarknet backbone network of YOLOv5, and the model structure is adapted and reconstructed. The experimental results show that MobileNetV3-Large achieves a good balance between detection accuracy and model complexity, with a 29% reduction in the number of parameters and a 42% reduction in FLOPs, which significantly improves the deployment efficiency of edge devices while maintaining better accuracy. This study provides a feasible technical path for the edge deployment of traffic sign detection algorithms in intelligent transportation systems.