Classification of Thermite Welding Defects using Local Binary Patterns and K Nearest Neighbors

Abstract

Radiography is a commonly used non-destructive method for inspecting thermite weld defects in welded rails. However, manual detection and classification of weld defects in radiographic images using human expertise remain a lengthy, costly, and subjective process. This process' success rate substantially depends on the inspector's ability to detect and classify defects. The development of an automated thermite weld defect detection and classification model will significantly improve railway infrastructure monitoring. This work proposes an image processing, and machine learning-based framework to automatically detect and classify thermite weld defects, in which the Chan-Vese Active Contour Model is used to define the Region of Interest and extract the weld joint. Features in the weld joint images are extracted using the Local Binary Patterns descriptor. The extracted features are then used to train a K Nearest Neighbor classifier. The proposed method achieved an average classification accuracy of 94 %.