Helmet detection algorithm of electric bicycle riders in urban traffic based on improved YOLOv8

Wearing a helmet as a basic safety regulation is crucial for the personal safety of electric bicycle riders during the ride. In practical situations, some riders neglect to wear helmets, which may lead to a greater risk of injury for riders in traffic accidents. Due to limited manpower, it is not possible to supervise and remind riders to wear helmets promptly and effectively. Therefore, to enhance the safety awareness of electric bicycle riders, this paper proposed a helmet-wearing detection framework based on You Only Look Once (YOLO) v8. The system detects whether electric bicycle riders wear helmets in real-time and promptly reminds them to take appropriate safety measures while driving. First, the most streamlined model in the YOLOv8 series, YOLOv8n, was chosen to train the detection model, thereby minimizing the parameters and calculations. Second, the Efficient Channel Attention (ECA) was integrated into the backbone to enhance the accuracy with a slight parameter increase. Experimental results demonstrated that the proposed algorithm achieved 80.9% accuracy with only five parameter growth, increasing by 1.2% compared to the original algorithm.