Abstract
As one of the popular techniques for non-destructive evaluation, infrared thermography often requires the assistance of data analysis models to help defect detection and identification. A novel generative independent component thermography (GICT) framework for defect detection in polymer composites is proposed. It utilizes a deep convolutional generative adversarial network to generate more informative images, which enhances the diversity of thermography data. The defect detection performance of sequential ICT-based thermographic data analysis can be enhanced. The feasibility of GICT is illustrated with its application to the defect detection of a carbon fiber reinforced polymer specimen.