GLCM and ANN based Approach for Classification of Radiographics Weld Images

Abstract

The process of welding involves welding defects. Welded material should be inspected accurately in order to ensure the quality of the design and operation. Non - Destructive Inspection is one of the important aspects which is responsible for identifying the flaw defect. An attempt has been made in the present work to accurately identify and classify the weld defects. A database of 79 images with 08 defects have been collected from Department of Mechanical and Industrial Engineering, Indian Institute of Technology Roorkee. The image database has been pre-processed and the features have been extracted by GLCM and feed to Artificial neural network for classification. Both 08 and 64 level features have been extracted by GLCM and fed to neural network. The features have been fed to both Feed Forward and Cascade Forward neural network for classification. Even though the quality of image database is not good, classification accuracy of 88.6% is obtained. © 2018 IEEE.