YOLO11-DAL a multi-scale fusion network for target detection in UAV-based power line inspection

Abstract

Abstract Reliable detection of transmission-line components and defects in unmanned aerial vehicle (UAV) imagery supports preventive maintenance. Repeated downsampling, however, removes fine spatial detail, while limited interaction between feature levels hinders small-object detection. We developed YOLO11-DAL, a compact detector based on YOLO11n, to address these limitations. C3k2-LEGM integrates local structure with global context in two lower-backbone blocks. Attentional Scale Sequence Fusion (ASF) aggregates P2–P5 features, while a Dynamic Aligned Head (DAH) coordinates classification and localisation. On a seven-class UAV power-line dataset, YOLO11-DAL achieved 86.2% mAP50 and 57.7% mAP50:95. These values exceeded the YOLO11n baseline by 3.3 and 2.5% points, respectively. The parameter count decreased from 2.58 million to 2.34 million. A prototype using a ZED camera and Jetson TX2 integrated image acquisition, inference and result visualisation. End-to-end latency was not measured, so the prototype establishes functional integration rather than verified real-time performance. The results support coordinated feature fusion and task-specific prediction as a practical design for compact UAV inspection detectors.