As a crucial area for mineral resource extraction, open-pit mining areas often face challenges including resource overexploitation, environmental damage, and safety hazards. Remote sensing object detection has emerged as a critical tool for dynamic monitoring in mining areas, though existing object detection algorithms demonstrate insufficient adaptability to the characteristics of remote sensing images such as diverse object scales, small targets, dense distributions, and complex backgrounds. This study establishes the GIOM dataset focusing on open-pit mining area and proposes YOLO-MAOD, a lightweight object detection algorithm for mining area surface features. To address the shortcomings of the YOLO11-s model in small object feature extraction and multi-scale detection, we propose the RAAtt module integrating region-aware concepts with SE channel attention mechanisms to enhance local feature capture capabilities. We introduce the CARAFE upsampling method to refine detail reconstruction and introduce the C2fPSA structure to optimize feature fusion. Innovatively proposing the CSPFasterDPF module, cross-level feature aggregation is achieved through efficient convolution operations to enhance multi-scale detection robustness. Experimental results show that the YOLO-MAOD model reduces parameters to 8.5M, achieving 87% and 55.9% :0.95 on the GIOM dataset, representing 2.6% and 3% improvements over YOLOv8-s respectively. Compared with YOLO11-s, it shows 1.8% and 1.4% enhancements while reducing parameters by 0.9M. The algorithm significantly improves detection accuracy in complex ‌open-pit mining‌ scenarios while maintaining lightweight advantages.