Region Grow Using Fuzzy Automated Seed Selection for Weld Defect Segmentation in X-Radiography Image

Abstract

To ensure the safety and the quality of weldment area during the welding process on products, it is essential to visualize the weld defects using the machine vision based inspection techniques. Commonly contrast of the x-radiography images are in low, ambiguous defects and absence of fine details. Since, segmenting and detecting the defects on weld x-radiography images is a thought-provoking task in Non-Destructive Testing (NDT) and it also depends on selection of suitable threshold. Seeded region grow is one of the most widely used method in image segmentation where seed selection is a challenging task. This paper presents a novel method, fuzzy automatic seed selection (FASS) in region grow based segmentation on weld x-radiography image. Otsu and Kittler threshold techniques based on the gray-level profile of an image addresses the automatic fuzzy seed selection procedure. The traditional methods such as Otsu and Kittler thresholds and proposed FASS methods are applied on the set of weld x-radiography images. The outputs are confirmed the performance of the proposed method.