Deep Learning-Enhanced Autonomous Drone System for High-Precision Aircraft Inspection

General visual inspections are imperative for commercial and military aircraft to identify damage that may imperil flight safety. Currently, these inspections are predominantly carried out by skilled maintenance personnel who meticulously inspect the aircraft’s surface to detect and document defects such as cracks, dents, corrosion, and broken fasteners. This manual process is time-intensive, prone to human error, and hazardous. Consequently, the implementation of computer vision and deep learning techniques for the automated detection of cracks, dents and corrosion in aircraft fuselage through advanced image processing techniques has gained traction across various fields. This advancement is facilitated by Unmanned aerial vehicles (UAVs) equipped with high-resolution cameras, where data is captured and processed utilizing sophisticated algorithms to enhance inspection accuracy and efficiency. The study utilized a dual-method approach, incorporating the Roboflow platform alongside Convolutional Neural Networks (CNNs) such as YOLOv8 and EfficientNet-B7, in addition to Canny Edge Detection to facilitate highly precise crack pattern identification and analysis. By utilizing a dataset of more than 2530 images, this sophisticated framework significantly enhances detection accuracy, achieving approximately 88.58%even in structurally complex and challenging-to-access regions. This research demonstrates the effectiveness of integrating MATLAB-based image processing with deep learning techniques to establish a robust model for aircraft crack detection, offering valuable insights for aviation maintenance and safety.