Factor analysis thermography for defect detection of panel paintings

ABSTRACT Active infrared thermography is an important non-destructive testing method used for revealing defect structures in materials. In many applications, thermographic data processing is necessary to extract defect features from a large number of thermal images. This work proposes to use a factor analysis thermography (FAT) method that automatically extracts defect features from thermograms via exploratory factor analysis, in tandem with a fuzzy c-means (FCM) clustering algorithm to segment the defects and background. By means of factor rotation, factor analysis minimises the complexity of factor loadings and makes the results more interpretable. Consequently, the defect information is extracted while large signal-to-noise ratios are obtained. Employing the FCM image segmentation algorithm on factor loading images reduces the interference of background on human visual detection. Additionally, the parameter selection is emphasised and addressed. Experiments on a panel painting illustrate that the proposed method promotes the accuracy and efficiency of thermographic detection of defects, compared with the popular principal component thermography (PCT) method.