Unmanned Aerial Vehicle Road Crack Detection Method Based on Improved YOLO11

With the wide application of UAVs in road inspection, target detection methods based on deep learning have become an important means to solve the problem of road crack detection. In order to improve the detection accuracy of UAV road inspection, this paper proposes an improved model RC-YOLO for UAV aerial photography scenarios based on YOLO11n. The improved model redesigns the feature dispersion pyramid network (FD-FPNet) based on the DASI module, which significantly enhances the model's multi-scale fusion capability for road crack features. In addition, the model proposes a lightweight detection head (LDH) that reduces the model complexity through group normalization and shared convolution, and improves the model's ability to localize and extract road crack features in UAVs. We validate the model on the UAV-PDD2023 dataset, and the results show that the improved model achieves 85.07% precision, 75.52% recall, 80.39% mAP50, and 52.19% mAP50-95, which are significantly better than the original YOLO model. The experimental results demonstrate that RC-YOLO has stronger detection performance and can be effectively applied to road crack detection under UAV perspective.