Not wearing or improper wearing helmets pose a significant threat to e-bike riders' safety,while manual check suffers from high workloads,low efficiency,and challenges in achieving full-area and all-time coverage.This paper proposes a helmet wearing detection method for riders based on an improved YOLOv5 model,which intelligently detects and automatically recognizes the helmet compliance through surveillance cameras.First,a dataset of e-bike riders' helmet wearing encompassing diverse locations,perspectives,weather conditions,and time periods is constructed to lay the research foundation.Subsequently,a helmet wearing detection model based on improved YOLOv5 is proposed,which improves the multi-scale feature fusion module of YOLOv5 to enhance the small-target detection effect.The Efficient Channel Attention (ECA) mechanism is introduced to strengthen feature map fusion performance,which significantly boosts model detection accuracy;a lightweight Neck is designed with GSConv to reduce detection time.Experiment results show that the proposed approach exhibits superior detection performance for helmet wearing of riders,achieving 93.2％ mAP (1.9 percentage points higher than YOLOX) with a detection speed of 15.23 ms per image.This balance of high detection rate and accuracy highlights its practical value for intelligent helmet wearing detection.