A new UAV-based approach for detection of paint damage on aircraft surfaces

In aircraft maintenance operations, the early detection of defects such as cracks, corrosion, impact damage and paint damage on aircraft external surfaces is of critical importance. Traditional visual inspection methods have significant limitations, including lengthy inspection times, high operational costs and results that vary depending on the operator. In this study, an approach combining Unmanned Aerial Vehicle (UAV)-based imaging with deep learning-based image processing techniques is proposed to address these limitations. As part of the study, a dataset comprising high-resolution images of aircraft external surfaces was created, with the aim of automatically detecting paint damage. YOLO-based object detection models (YOLOv11 and YOLO26) were used for defect detection, and model performance was evaluated comparatively through training conducted on the same dataset using the Roboflow and Google Colab platforms. The results obtained demonstrate that the proposed approach provides high accuracy and reliability in the detection of paint damage on aircraft surfaces. Furthermore, it was observed that the choice of training platform influences the model’s temporal consistency in video-based analyses. Tests conducted on different hardware platforms demonstrate that the method is applicable across various systems; however, hardware capacity is a critical factor affecting real-time performance. In conclusion, the proposed UAV-supported deep learning approach offers a faster, non-contact and repeatable inspection method for aircraft maintenance processes, demonstrating significant potential for digitalisation and automation in Maintenance, Repair and Overhaul (MRO) operations. © 2026 The Author(s).