Core ML-Based Aircraft Skin Damage Detection System Using YOLO11

This study introduces a software that was designed for aircraft skin damage inspections using portable, lightweight devices, such as iOS devices while combining it with Unmanned Aerial Vehicles (UAVs) application to allow aerial image inspection in real-time. Our objective is to utilize the computational capabilities of these devices to perform inspections seamlessly without the reliance on external devices in which eventually aimed at aiding the routine inspections for airworthiness and identifying potential risk upon flight. By using the object detection algorithm YOLO11, combining with the iOS-based Core ML algorithm, we created a software that can perform detection on aircraft skin defect and immediate return results in real-time. With additional functions, allowing users to perform inspections using UAV and thermography input. The experimental results confirmed the effective recognition of defects with the mean average precision (mAP@0.5) at 0.847 for Yolo11 nano (n) for all classes using visible light, and an mAP@0.5 at 0.662 when including thermography detection method. However, despite the effective detection, problems, such as the low in mAP when utilizing thermography detection, the effect of heat on the test device, and the test device that reported overheating during the operation, greatly affect the reliability during the inspection. While there is still room for enhancement, this study highlights the potential of implementing automated image detection technology to support inspection processes within the aviation industry.