Defect Detection in C-scan Maps

An image processing procedure is proposed aiming at detecting porosity defects in materials used in the aerospace industry by means of the analysis of C-scan images obtained by ultrasound inspection techniques. An image is described through a number of features, which are then analyzed in order to evaluate the similarity of an unknown examined image with respect to a set of reference ones. The procedure is performed by firstly applying a 2D wavelet transform of the input image and then a feature extraction process based on statistics of the detail images produced by the transform itself. A data reduction process is then applied using the principal component analysis technique. Finally, a decision ruler performs the image classification by considering the distance among points in the principal component plane. An application of this procedure is presented for the analysis of single images, giving a localization of defects. Preliminary results, obtained by processing both simulation images and real C-scan maps, showed the efficacy of the proposed method.