Subsurface Defect Detection in Carbon Fiber-Reinforced Plastics Using Infrared Polarization Imaging

Carbon fiber-reinforced plastics (CFRP) are widely used in various industries due to their high strength and light weight. However, subsurface detects can significantly reduce the material's lifespan and compromise safety. Many existing nondestructive testing methods often struggle with challenges such as poor detection of various defects, susceptibility to noise, and surface interference. This paper presents a robust technique for subsurface defect detection in CFRP, combining a future frame prediction network with infrared polarization imaging that captures both thermal and polarization data. The future frame prediction network is designed to extract physically meaningful spatio-temporal features of detects based on thermal diffusion, while polarization data serves as auxiliary information to reduce surface interference. The proposed method is validated on a newly collected CFRP polarization defect dataset. Experiment results demonstrate that the proposed approach outperforms existing thermography-based techniques, particularly in complex conditions involving uneven heating and diverse surface interferences. Experiment results demonstrate that, under non-uniform heating and surface interference conditions, the future frame prediction-based defect detection method achieves a 3.24% improvement over existing thermography-based approaches. Incorporating polarization information further enhances performance by 5.18%, while maintaining a lightweight architecture with only 2.19M parameters and a relatively fast inference speed.