Self-guided filtering slow feature thermography for subsurface defect detection in composite materials

Abstract

Infrared thermography (IRT) is widely used for non-destructive measurement and testing of composite materials and heritage conservation. Thermal data post-processing techniques playing a crucial role in improving accuracy and efficiency. In this work, a self-guided filtering-based slow feature thermography (GF-SFT) method is proposed for subsurface defect assessment of composites. It processes and analyzes time-series thermal imaging data of inspection targets from both temporal and spatial dimensions. First, a self-guided filtering method enhances the thermal image sequence in the spatial domain, preserving critical image features (e.g., edges) while reducing noise and suppressing irrelevant details, thereby improving the data quality. Subsequently, the slow feature analysis technique isolates stable background components in the thermal signal over time while distinguishing rapidly changing features associated with subsurface defects. The proposed framework is validated through an evaluation of filtering performance and a Fourier-based interpretation of fast- and slow-varying thermal features. A test case on a carbon fiber reinforced polymer demonstrates the effectiveness of the GF-SFT method in detecting subsurface defects.