Based on Multi-Module Synergistic Optimization: MSDA-YOLO Method for Foreign Object Detection on Power Transmission Lines

Foreign object detection on power transmission lines using UAVs faces challenges such as multi-scale feature discrepancies, morphological variability, sub-optimal fusion of low-level and high-level features, and strong background imbalance. To address these issues, we propose MSDA-YOLO, an enhanced YOLOv11n-based detector that integrates four task-specific modules: a multi-scale morphology-aware attention (MSMA) module for small and elongated objects, an adaptive dual-branch fusion and alignment (ADFA) neck for cross-level feature aggregation and background suppression, a morphology-aware anchor design (MAAD) module that generates shape-matched anchors, and a morphology-adaptive distance and width–angle loss (MADWL) for stable regression. Together they form a coherent pipeline of feature enhancement, alignment fusion, anchor matching and regression optimization while preserving real-time inference.The model is evaluated on a self-built dataset of 4,520 single-modal UAV RGB images containing bird nests, balloons, kites and trash. MSDA-YOLO achieves an mAP50–95 of 74.7%, exceeding the YOLOv11n baseline by 3.0 percentage points. For small targets such as kites and trash, the mAP50–95 increases by 3.4% and 4.8%, respectively. With 3.109 million parameters and 6.7 GFLOPs, MSDA-YOLO satisfies the real-time requirements of UAV edge devices and provides a practical solution for large-scale power grid inspection.