基于ME-YOLO11的工人不安全行为图像检测算法OA

In industrial production and construction sites,unsafe worker behaviors such as not wearing safety helmets,smoking,and using mobile phones are major causes of safety accidents,and thus require accurate recognition and real-time monitoring.However,due to the small size of targets like helmets,cigarettes,and phones,the existing YOLO 11 ob-ject detection algorithm suffers from missed and false detections.The core reasons lie in the limited multi-scale feature representation capability of its backbone and information loss during cross-layer feature fusion in the neck.To address these issues,this paper proposes an improved algorithm,ME-YOLO11.A new feature processing module,C3k2_MLCA,is introduced to replace the original C3k2 in the backbone,enhancing the model's ability to extract features from small objects.Additionally,an enhanced feature fusion pyramid network(EFFPN)is designed in the neck,incorporating DySample dynamic upsampling,a content-guided attention fusion layer,and a micro-target fusion layer to strengthen multi-scale fea-ture interaction.Experimental results show that ME-YOLO11 achieves 91.2％mAP@0.5 and 64.0％mAP@0.5:0.95 on the self-built WUBD dataset,and 38.5％and 23.2％respectively on the VisDrone2019 dataset,demonstrating excellent performance in small object detection and strong generalization capability.