Feeling weak, eyes squinting, and unbearably tired. New products keep rolling down the conveyor belt for inspection. The cornucopia keeps giving. Every single product must be meticulously examined. Mistakes are not allowed. It’s exhausting, but what if quality assurance could be made easier?
The uniform quality of products is of utmost importance to manufacturing industries. Yet, deficiencies in the precision and systematic approach of quality assurance on production lines are surprisingly common. Typically, a large amount of quality data must be collected, which opens the door to human error. Maintaining this data can also stagnate without proper tools. Customer complaints — and, in the worst-case scenario, recalls of substandard products already on the market — are costly, both financially and in terms of the company’s reputation.
How can quality assurance be made precise and effortless?
Vogel (2022) suggests digitizing quality assurance processes to improve efficiency in meeting quality standards. One method is the use of image recognition, where computers are trained to interpret visual data by utilizing computer vision applications and image recognition techniques (Shaip 2024). With image recognition, defects or flaws in products manufactured on the production line can be detected, such as knots, cracks, and holes in wooden panels. Centria University of Applied Sciences’ cybersecurity team has studied image recognition and compared the performance of deep learning models — MobileNet, MobileNetV2, ResNet-50, and VGG-19 — in detecting quality deviations in wooden panels (Tuunainen, Isohanni & Jose 2024).
Do you want to know more? Read (in Finnish) about this topic in Centria Bulletin: https://centriabulletin.fi/tekoalysta-apua-laadunvarmistukseen/
Centria Bulletin (ISSN 2489-3714) is an online journal where the article was first published.
Tom Tuunainen
R&D Developer
Centria University of Applied Sciences
Tel. +358 40 681 7207
References
Vogel, M. 2022. Digitalisaatioon sopivaksi: laadunhallinnan kolme vahvuutta. DQS Holding GmbH. Available at: https://www.dqsglobal.com/fi-fi/opi/blogi/digitalisaatioon-sopivaksi-laadunhallinnan-kolme-vahvuutta
Shaip. 2024. Mikä on AI-kuvantunnistus? Miten se toimii ja esimerkkejä. Available at: https://fi.shaip.com/blog/what-is-ai-image-recognition-and-how-does-it-work/
Tuunainen, T., Isohanni, O., & Jose, M. R. 2024. A comparative study on the application of Convolutional Neural Networks for wooden panel defect detection. IEEE 22nd World Symposium on Applied Machine Intelligence and Informatics (SAMI). Stará Lesná, Slovakia, 321-326. Available at: https://doi.org/10.1109/SAMI60510.2024.10432810