基于深度学习的煤矿井下人员检测算法OA

Personnel detection is an important aspect of ensuring coal mine safety production and constructing intelligent mines.Given the complex underground environment of coal mines and the difficulties in personnel detection,an improved YOLOv8-F personnel detection method for coal mine underground environments is proposed.The MobileNetV4 efficient network is integrated into the YOLOv8 backbone to improve accuracy while reducing computational cost.To ensure thorough feature extraction,a DAttention mechanism fused with DCNV4 is added,improving the accuracy and completeness of feature information.Finally,by introducing a boundary loss function,the precision of the bounding boxes is improved,achieving accurate personnel detection.Experimental results show that,on a specific dataset of coal mine workers'actions,compared with the baseline model YOLOv8n,the YOLOv8-F model's accuracy,recall,and mAP improved by 3.2,2.7,and 2.3 percentage points,respectively,while the floating point operations(GFLOPs)decreased by 1.4 G.The improved model achieves a great balance between detection accuracy and model lightweightness,verifying the effectiveness of the new algorithm.