A Fire Detection Method based on Improved YOLOv8 in Automatic Coal Sample Transport

Large domestic energy ports have strict requirements for safety. The precise detection of early fire during the automatic coal sample transport, can significantly contribute to preventing the occurrence of unsafe incidents and mitigating potential hazards. In view of the small size of early-stage flames and smoke, as well as their complex deformation, which easily leads to missed detections and inaccurate localization, this paper proposes a fire detection method based on the improved YOLOv8. Firstly, a deformable small object branch(DSOB) based on deformable convolution is proposed to improve the detection accuracy of tiny flames and smoke. Secondly, Wise Intersection over Union(WIoU) loss function is used instead of the Complete Intersection over Union(CIoU) loss function, which avoids the negative impact of low-quality bounding boxes of the deformable objects on model training. Experimental results on the fire dataset show that the fire detection method based on our improved YOLOv8 can significantly achieve accurate detection of fire, outperforming other methods in terms of mean average precision (mAP).