Crack Detection on Aircraft Composite Structures Using Faster R-CNN

Detecting imperfection and cracks of composite internal surfaces on aircraft is an arduous task due to accessibility problem, dark color surface, limited space and poor lightings. In this paper, we present a deep learning framework based on Faster R-CNN for crack detection on the internal composite surfaces of civil transport aircraft. To enhance recognition efficiency, a Faster R-CNN network is designed. To overcome the constraint of limited sample images, image augmentation method is developed. Experiment shows that the overall performance and robustness of the proposed system could meet our project requirements.