Vehicle Embedded Traffic Sign Recognition

In response to the challenges encountered in traffic sign detection and recognition, such as small target size, variable shape, difficulty in feature extraction, and susceptibility to complex background, this study proposed a lightweight network model called SCF-YOLOv5. The algorithm integrates the channel attention mechanism before the SPPF layer of the backbone network of the YOLOv5 model, which enhances the model's ability to identify key information. In the neck network, a lightweight general up-sampling operator CARAFE was used instead of traditional up-sampling techniques, improving image resolution through content-aware technology. Furthermore, the CIOU loss function was optimized to Focal loss, effectively addressing class imbalance and sample imbalance issues. Finally, the algorithm was deployed on the Raspberry PI embedded platform. Compared with YOLOv5s, the number of parameters decreased by 40.0%, FPS increased by 11.05%, and mean average precision increased by 3.4%, which enhanced the accuracy and robustness of the algorithm.