AI-Based Detection, Analysis, and Classification of Aircraft Surface Defects Using Deep Learning Hava Araci Yüzeylerindeki Kusurlarin Derin Öǧrenme ile Yapay Zeka Tabanli Tespiti, Analizi ve Siniflandirmasi
34th Signal Processing and Communications Applications Conference, SIU 2026, İstanbul, Turkey, 7 - 10 July 2026, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/siu71813.2026.11636618
- City: İstanbul
- Country: Turkey
- Keywords: Aircraft Surface Damage, Computer Vision, Deep Learning, Defect Classification, Image Processing
- Yıldız Technical University Affiliated: Yes
Abstract
Early detection of aircraft surface defects is important for maintenance reliability. This study proposes an AI-based system for the automatic detection and classification of five defect types - dents, cracks, paint peeling, scratches, and missing-head - using 4,557 images. The system employs a hybrid architecture combining EfficientNet-based CNN feature extraction with K-means clustering. Comparative results show that the supervised approach achieves higher performance across all metrics. Accuracy and recall improve by approximately %76, F1-score by %68, and precision by %48, while the F1-score increases from 0.5492 to 0.9266. The proposed system serves as a decision-support tool aimed at reducing error rates and accelerating inspection processes.