Inspection of manufacturing processes has become an essential part of industry 4.0 nowadays. Quality assessment at each step of production enables the detection of flaws at early fabrication stages, reducing materials usage, thus cutting manufacturing costs.
Furthermore, it mitigates the risk of defects appearing in sold production. Non-invasive check-ups, such as those computer vision (CV) based, might be suitable for most observable defects, as the visual-based approach is routinely employed for defect
detection in the industry.
While accurate for simple defects, human-performed visual inspections remain monotonous, time-consuming, and susceptible to fatigue-caused errors. In addition, they can create bottlenecks in the manufacturing pace as more personnel is needed to
keep up with massive production. Human inspection becomes particularly problematic in more complex defects as subjectivity surges, resulting in inconsistent inspection accuracy error rates. In contrast, automated CV systems can perform defect inspections continuously, faster, and more accurately than humans. CV systems also have the advantage of being scalable with computation resources. Nevertheless, automated vision-based systems still face various challenges, namely in distinguishing defective surface patterns or being adaptable to changes in the production line. Even with the existent spectrum of approaches to engaging conventional CV algorithms for various tasks, constraining sophisticated visual data dynamics under predefined rules is still challenging.
Recently, there has been a significant rise in interest in employing deep learning for visual understanding.
Project funding:
KTU Research and Innovation Fund
Project results:
System for detecting visual defects in the manufacture of patterned furniture panels.
Period of project implementation: 2022-04-22 - 2022-12-31
Project coordinator: Kaunas University of Technology
