Slow Feature Thermography for Non-Destructive Evaluation of Defects in Composite Materials

In recent years, various methods for processing and analyzing thermographic data have been applied in non-destructive testing to improve the visibility of defects. Representatively, traditional linear dimensionality reduction methods (e.g. principal component analysis) have achieved good results in thermographic data analysis. In this work, a slow feature thermography (SFT) method for non-destructive evaluation of defects in composite materials. Thermographic data can be regarded as a continuous time series generated by an infrared imager at different moments. Slow feature analysis algorithm is subsequently used to extract slow features of the thermal image sequence. Slow features refer to features that are unchanged or slowly changing in time-series data, such as the background of a thermal image. By extracting the slow features of data, the inhomogeneous background, noise and defect features contained in thermal images can be effectively distinguished. Test results on carbon fiber reinforced polymer specimens demonstrate the feasibility and effectiveness of the SFT method.