RMU-YOLO: A Deep Learning-Based Algorithm for Aircraft Skin Damage Detection

To address the challenges of multi-scale variation, complex backgrounds, and lightweight model deployment in aircraft skin damage detection, an improved YOLOv11n algorithm, named RMU-YOLO, is proposed. This algorithm integrates a RepGFPN feature pyramid network to enhance multiscale feature extraction capabilities and combines an MSDA (Multi-scale Dilated Attention) mechanism with a C2PSA module to suppress background noise and strengthen key features. Furthermore, lightweighting is achieved through the use of UIB and C3k2 modules, and the Unified-IoU loss function is employed to optimize bounding box regression. Experimental results on a self-built dataset demonstrate that, compared to YOLOv11n, RMU-YOLO achieves higher detection accuracy and efficiency. The [email protected] is improved by 3.4%, precision and recall are increased by 3.3% and 2.9% respectively. It reduces the computational load compared to the baseline, which in turn boosts inference speed, striking a superior balance between computational cost and detection performance., providing a reliable solution for automated detection in this field.