Automated Aircraft Structural Defect Detection Using Deep Learning and Computer Vision

Manual aircraft inspections are labor-intensive and susceptible to human error, potentially compromising safety and accuracy. This study presents an automated defect detection framework based on the Mask R-CNN instance segmentation model for identifying cracks and dents in aircraft structures. A dataset of 2,000 annotated images was generated using augmentation techniques and used to train a ResNet-101-based Mask R-CNN model. The system achieved high detection performance, with crack detection reaching a precision of 92.8%, recall of 88.7%, and F1 score of 90.75%; dent detection achieved 91.2% precision, 88.1% recall, and an F1 score of 89.62%. Evaluation using mean Intersection over Union (IoU) and Average Precision (AP@[IoU=0.50:0.95]) confirmed accurate defect localization and segmentation. These findings demonstrate the model's potential to improve inspection reliability and operational efficiency, contributing to safer, more consistent aircraft maintenance practices.