基于ESH-YOLO的小目标安全帽检测算法OA

Aiming at the problems of low detection accuracy,false detection and missed detection of existing safety helmet wear-ing detection algorithms for small targets in complex scenes,this paper proposes an improved safety helmet wearing detection al-gorithm ESH-YOLO based on YOLOv8n.Firstly,the Star_CAA module is obtained by combining the Star Blocks network with the CAA attention mechanism,and the module is used to optimize the C2f structure in the original network backbone to enhance the recognition performance of the model for safety helmets in complex scene images,giving more attention weight to small target safety helmets and inhibiting the attention to background and worker regions.Secondly,ELA_HSPAN is used as the neck struc-ture.Through feature selection and fusion of feature maps with different resolutions,the feature information of each level is better relayed,and the detection and localization ability of objects with different scales,especially small objects,is improved.Finally,the PIoUv2 loss function is introduced to accelerate the convergence speed of model training and improve the positioning accu-racy of bounding box regression.Experimental results show that compared with the original model,the ESH-YOLO model im-proves the precision,recall rate,mAP@0.5 and F1 score by 0.9 percentage points,3.6 percentage points,2.8 percentage points and 2.3 percentage points respectively,which can effectively reduce the missed detection and false detection of small targets in complex scenes.The application of this algorithm in complex construction sites can accurately detect whether workers wear safety helmets,improve the wearing rate of safety helmets,so as to ensure the life safety of workers.