Improvement of defect detection in shearography by using principal component analysis

A post-processing technique based on principal components analysis (PCA) is proposed for shearography for defect detection. PCA allows decomposing a time series of images into a set of images called Empirical Orthogonal Functions (EOF), each showing features with a given variability in the time series. We have applied PCA on composite samples containing various defects at different depths and which undergo transient thermal wave. Analyzing the temporal series shows the shallow defects appearing first whereas the deeper ones appear later. With PCA all the defects appear in one or two of the EOF, easing the identification of defects.