Exogenous fires in underground coal mines are characterized by low illumination, smoke occlusion, heavy dust loading and pseudo fire sources, which jointly degrade image quality and cause missed and false alarms in visual detection. To achieve accurate and real-time early warning under such conditions, this paper proposes a mine exogenous fire detection algorithm based on an improved YOLOv9m, termed PPL-YOLO-F-C. First, a lightweight PP-LCNet backbone is embedded into YOLOv9m to reduce the number of parameters and GFLOPs while maintaining multi-scale feature representation suitable for deployment on resource-constrained edge devices. Second, a Fully Connected Attention (FCAttention) module is introduced to perform fine-grained frequency–channel calibration, enhancing discriminative flame and smoke features and suppressing low-frequency background clutter and non-flame textures. Third, the original upsampling operators in the neck are replaced by the CARAFE content-aware dynamic upsampler to recover blurred flame contours and tenuous smoke edges and to strengthen small-object perception. In addition, an MPDIoU-based bounding-box regression loss is adopted to improve geometric sensitivity and localization accuracy for small fire spots. Experiments on a self-constructed mine fire image dataset comprising 3000 samples show that the proposed PPL-YOLO-F-C model achieves a precision of 97.36%, a recall of 84.91%, mAP@50 of 96.49% and mAP@50:95 of 76.6%, outperforming Faster R-CNN, YOLOv5m, YOLOv7 and YOLOv8m while using fewer parameters and lower computational cost. The results demonstrate that the proposed algorithm provides a robust and efficient solution for real-time exogenous fire detection and edge deployment in complex underground mine environments.

To address equipment blockage and belt damage caused by large coal blocks on conveyor belts, this study proposes SCCG-YOLO, a lightweight real-time detection model based on YOLOv8n. The model introduces CPNGhost into the backbone to enhance receptive-field coverage and edge-detail extraction for large targets, incorporates Shuffle Attention in feature fusion to improve discriminability under complex lighting and dust interference, replaces fixed upsampling in the neck with CARAFE to refine high-level semantic reconstruction, and adopts DIoU loss to strengthen geometric constraints during bounding-box regression. Experiments were conducted on a task-specific derivative subset of the public CUMT-Belt dataset. After image screening, label correction, and re-annotation, 1,276 valid images were retained and divided into training, validation, and test sets at a ratio of 8:1:1. The results show that SCCG-YOLO achieves 91.9% mAP@50, 532.6 FPS, and only 2.7 MB parameters, demonstrating a favorable balance among detection accuracy, efficiency, and model compactness. These results indicate that the proposed method can satisfy the real-time detection requirements of underground conveyor-belt operation and has practical value for intelligent mine safety warning.