Safety Helmet Wearing Detection in Aerial Images Using Improved YOLOv4

In construction, it is important to check whether workers wear safety helmets in real time. We proposed using an unmanned aerial vehicle (UAV) monitor construction As the small target of photography poses challenges safety-helmet-wearing detection, we improved YOLOv4 model detect helmet-wearing condition photography: (1) By increasing dimension effective feature layer backbone network, model's receptive field reduced, and utilization rate fine-grained features improved. (2) introducing cross stage partial (CSP) structure into path aggregation network (PANet), calculation amount efficiency at different scales (3) The complexity reduced by group convolution pruning PANet multi-scale detection mode for de-redundancy. Experimental results show that achieved highest performance UAV helmet task, mean average precision (mAP) increased from 83.67% original 91.03%, parameter 24.7%. prove can effectively respond requirements real-time wearing photography.