Safety Equipment Wearing Detection Based on YOLO-MT in Coal Yards

Coal workers wear safety gear at the coal yard can greatly improve the personal safety of coal miners. But workers have weak safety awareness, often forget to wear safety equipment. To ensure production safety, a coal yard workers' safety equipment wearing detection problem is considered in this paper, and an improved YOLO-MT detection model was presented to address the issue of low detection accuracy of existing object detection algorithms in coal yard environment. Firstly, a neck network structure of YOLOv8 was strengthened by texture information to help the model better absorb texture information and enhance its positioning capabilities. Then, a macro residual backbone is created as the detection algorithm's backbone network to enhance the network's detection performance. Thirdly, the WMPDIoU loss function was created to enhance the network's regression efficiency and detection accuracy by integrating the benefits of the MPDIoU and WIoU loss functions. According to experimental data, the YOLO-MT algorithm improves [email protected] by 2.5% when compared to the baseline model, and it can achieve an FPS of 203.2, which is sufficient for real-time detection.