The safety of electric vehicle (EV) riders has become a major issue in society, and wearing safety helmets has been proven to effectively reduce injuries in accidents. To enhance road traffic safety, improve regulatory efficiency, and reduce the labor cost of round-the-clock inspections, we propose a fast and intelligent helmet detection algorithm with UAV assistance based on deep learning. Firstly, an improved Outlook-C2f architecture was proposed to enhance the algorithm’s focus on the small targets; Secondly, we proposed using CARAFE in the Feature Pyramid Network (FPN) to dynamically generate weights for precise feature reconstruction and improved spatial resolution; Finally, Wise Intersection over Union (WIoU) was integrated to improve the accuracy of positional information. Experimental results show that the improved YOLOv8 algorithm achieved a mean Average Precision (mAP) of 96.7%, outperforming YOLOv8 by 1.3%, and FPS of 26.91, offering a balance between accuracy and speed compared with other algorithms based on our dataset.