To achieve high-precision detection of helmet wearing in road scenarios,this paper proposes a multi-scale attention and spatial feature fusion network (MSASNet).Specifically,a multi-scale attentional feature fusion (MSAFF) module is designed to strengthen the model′s focus on key helmet regions under complex backgrounds;an adaptively spatial feature fusion (ASFF) module is introduced to enhance the effective integration of multi-level features,thereby improving detection accuracy; and a hybrid loss function is developed to optimize the detection of small-scale helmet targets.Experimental results demonstrate that MSASNet achieves a mean average precision (mAP) of 95.2% on the helmet-wearing detection dataset,which is 1.7 percentage points higher than YOLOv11,while maintaining a high-level real-time performance.This study provides technical support for helmet detection in complex road environments and shows significant potential for practical applications.