Multi - Class welding flaws classification using texture feature for radiographic images

Abstract

The paper presents a novel approach for multi-class weld flaw classification by means of Gray level co-occurrence matrix (GLCM) based texture feature extraction technique and Artificial Neural Network classifier. The weld radiography films have been digitized using CCD camera, followed by image processing techniques i.e. RGB to gray conversion, region of interest (ROI) selection, noise reduction and contrast enhancement. Subsequently a set of 8, 64 and 44 texture features vectors have been obtained from each of the digitized weld images by means of GLCM. Further, the features obtained have been classified with cascade-forward back propagation neural network. The proposed system has obtained overall classification accuracy of 86.10% for nine different types of weld flaws of digitized radiographic images. © 2014 IEEE.