Abstract: To exploit the technical merits of unmanned aerial vehicle (UAV) inspection in the urban rail transit industry and realize efficient adaptation between UAV aerial images and bridge deck crack detection models, this paper proposed a bridge deck segmentation and crack identification model based on YOLO11n-seg. The paper adopted YOLO11n-seg to generate binary mask images. It performed bitwise AND operations between the mask images and original aerial images to accurately extract bridge deck areas and eliminate background interference, implemented data augmentation on the segmented bridge deck datasets to enhance the generalization performance of the model, and then reused the YOLO11n-seg framework to precisely identify bridge deck cracks and improve the robustness of the model. Test results reveal that the proposed model achieves significantly improved convergence compared with ablation experiment groups. It delivers superior overall performance against mainstream crack detection models. Furthermore, the model accurately identifies single and multiple cracks under complex background disturbances and exhibits favorable robustness. Its inference speed reaches 0.16 s per image, which satisfies the demand for real-time result output during on-site UAV inspection.