Abstract: The prevention and control of spontaneous combustion in open-pit coal seams are crucial for ensuring the safety and efficiency of coal mining operations. Existing detection methods mainly rely on manual inspection and non-intelligent thermal imaging or gas detection instruments, which suffer from low real-time performance, low efficiency, and limited accuracy. To address these issues, this study proposes a UAV-based detection method for spontaneous combustion in open-pit coal seams, aiming to improve detection efficiency and economic benefits. To overcome challenges such as low contrast between infrared targets and backgrounds, small target sizes, and confusion between high-temperature exhaust from excavators and coal seam hot spots, a UAV-based Open-Pit Coal Seam Spontaneous Combustion dataset (UAV-OCST) was constructed. An improved YOLOv8n model, named YOLO-WRC, is proposed. The model introduces a C2f-WT module to optimize feature component partitioning, employs a re-parameterized generalized feature pyramid network to enhance feature extraction and fusion, incorporates a lightweight distributed focus detection head to integrate multi-level local and global information, and utilizes the PIoUv2 loss function to improve regression performance. Experimental results on the UAV-OCST dataset demonstrate that YOLO-WRC achieves significant improvements, reaching 95.4% mAP@50, 90.1% recall, and 88.2% detection precision, outperforming several mainstream models and verifying the effectiveness and superiority of the proposed method.