Weld classification using gray level co-occurrence matrix and local binary patterns

This paper presents an algorithm that can classify weld seams from images, exploiting machine learning techniques. Manual visual inspection is the primary way of evaluating weld seams, in cases where the primary goal is to keep inspection costs low. Such, visual inspections entail manual interpretation and evaluation, which are both time consuming and the result often depends on the person assigned to the task. These drawbacks render automatic visual inspection appealing. Thus, this paper seeks to find a possible solution for the visual inspection of welds, where two feature extraction methods are examined and tested in conjunction with two different classifiers. We investigate whether visual inspection based on texture-describing features, processed with a machine learning algorithm, can detect flaws and defects in a weld merely by inspecting the surface of the object, in a way similar to how human eyes detect them and we achieve 96% classification accuracy on a new dataset.