Abstract
Visual inspection is currently the most common way for surface defect detection in the damaged composite structure, which is considered unreliable for barely visible impact defects (BVIDs) of less than 0.3 mm in depth. This paper proposes a defect detection method using stereo vision-based surface reconstruction and curvature analysis, which can detect defects of less than 0.1 mm in depth. To meet the real-time requirement of on-site detection, a real-time image registration algorithm is proposed through deformation decomposition and deep learning-based image correlation, realizing 1 frame/s calculation speed for megapixel images. The detection experiments of barely visible defects on carbon fiber-reinforced plastic (CFRP) composite laminates are carried out to demonstrate the effectiveness and efficiency. The proposed method presents a low-cost and high-efficiency approach for non-destructive testing and health monitoring of composite structures.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have