Enhancement of Outdated Vision Systems in SMEs with Artificial Intelligence Powered Solutions

This paper deals with the application of Artificial Intelligence (AI) based software solutions to the quality assurance in manufacturing of the printed circuit boards (PCBs). We have used Convolutional Neural Network (CNN) and YOLOv4 for image defect detection at older vision systems that are based on predefined pixel detection. We realised three case studies of using AI with old camera setup and low computer resources, since it is often the case in real industrial systems: 1) simple image recognition, 2) recognition of the simple true or false object parameters and 3) more advanced recognition of certain defect types at the observed object. The results showed that it is possible to use AI/based solutions with older camera and computer setups, even though the time needed was significantly longer that in the case of up-to-date visual and computer resources. Application of an AI-based systems for image defect detection with older visual and computer systems reduced pseudo failures from 10% to 3%, thus making significant financial savings.