Enhanced YOLOv9 for Pedestrian and Vehicle Detection in Foggy Traffic Scenarios

In the development of autonomous driving technol-ogy, accurately detecting pedestrians and vehicles in foggy traffic scenarios is a critical challenge. In response to the challenges faced in traditional pedestrian and vehicle detection methods, which of-ten suffer from low precision and reliability, especially in foggy conditions, a modified YOLOv9 network model has been proposed. By integrating the Deeplabv3+ and the SE (Squeeze-and-Excitation) attention mechanism modules into the YOLOv9 model, the method harnesses the simultaneous capture and synthesis of ex-pansive and detailed features across the imagery, augmenting the model's proficiency in recognition and understanding. This en-sures the model's robustness and accuracy under foggy conditions. Experimental results indicate that the YOLOv9 model, after its enhancement and when evaluated on a specially processed VOC format dataset, saw its average precision rate reach 85.9%, marking a 16.5% improvement over the original YOLOv9 network. In foggy traffic conditions, the upgraded YOLOv9 outperforms con-ventional detection networks in tasks related to pedestrian and Vehicle identification.