APPLICATION OF YOLO NEURAL NETWORKS IN DEFECT DETECTION IN AUTOMATED QUALITY CONTROL SYSTEMS FOR PRODUCTS OF ORGANIC ORIGIN

Abstract

Modern technological equipment for various purposes must fully meet the requirements of digital production and be able to quickly integrate into the structure of smart enterprises that are switching to cyber-physical technological systems. Machine vision (MV) is a key element of such systems and a promising automation tool that enables the capture and movement of various objects, including components, quality control, and safety. In addition, machine vision is increasingly being used in modular machine tools. The article analyses various applications of machine vision, in particular, its use in intelligent technological systems for product quality control. Particular attention is paid to fast and efficient quality analysis at the stage of the production process, which allows for accurate defect detection. The article investigates the feasibility of using mathematical models of artificial neural networks to create an intelligent system for monitoring the geometric condition of products. The aim of the study is to identify and classify the quality parameters of products of organic origin, namely reed tubes. For this purpose, new quality control methods based on computer vision and machine learning algorithms are proposed to identify and classify various types of defects using an integrated approach, namely, technologies for localization and classification of product defects using neural networks. As an example, we consider products of organic origin, for which the localisation and classification of defects is difficult due to their natural structure and uniqueness. To solve this problem, the use of neural networks of several YOLO architectures was first proposed. The study presented the results of training two modifications of the YOLOv10s and YOLOv10m neural networks, which were positive.