Problems of Surface Defectoscopy of Metals using Machine Learning and Ways for Their Solutions

Rejection of metal products is an important stage of the production process aimed at ensuring the best quality of the final product. Traditional rejection methods, based on visual inspection or the use of simple automated systems, have their limitations and disadvantages, such as low speed and accuracy of defect classification. The paper examines the possibility of using various machine learning methods to classify defects in metal products. A comparative analysis of these algorithms, as well as their effectiveness, is carried out in order to determine the most suitable approach to the automatic rejection of metal products.