Abstract: Accurate detection of mining machinery in UAV remote sensing imagery is crucial for intelligent monitoring of open-pit mines. To address key challenges in identifying heterogeneous machinery targets in UAV-captured open-pit mine images—namely significant variations in target scales, complex background interference, and frequent missed detections of small targets—this paper proposes CDMG-YOLO, a structurally optimized lightweight remote sensing object detection model based on YOLOv11n. The model incorporates a P2 detection head, a D-C3k2 deformable large-kernel convolution module, a dual-attention mechanism combining CAFM and MCAM, and a GHM-based loss function to achieve high-precision detection of small targets and robust performance under complex backgrounds. Experiments on a self-built open-pit mine dataset and the publicly available LEVIR dataset show that CDMG-YOLO achieves a precision of 0.859, mAP@0.5 of 0.697, and mAP@0.5-0.95 of 0.547, outperforming current mainstream lightweight YOLO variants while maintaining a compact model size with only 3.1 million parameters and 12.7 GFLOPs. The results demonstrate the robustness and applicability of the proposed method for small-object detection in complex mining areas using remote sensing imagery..