Vision-Based Quality Control System for FMCG Production
Project Title: Vision-Based Quality Control System for FMCG Production
Client: Bunge Polska Sp. z o.o.
Industry: Life Science & Health
Expertise: Application Development
Scope of the project: Design and implementation of a system for label verification on production lines
Tools: Python, .NET, OpenCV, scikit-image
A label verification system for production lines is a solution that uses vision technology and AI algorithms to compare labels against approved templates in real time. It automatically detects errors such as typos, incorrect barcodes, or missing graphic elements, eliminating the need for manual inspection.
ALTEN Polska (formerly Solwit) delivered a vision-based quality control system for Bunge Polska, a producer of vegetable oils. The system was designed to automate and enhance the precision of label verification during production.
The expertise of ALTEN Polska’s team included leveraging Computer Vision algorithms, enabling rapid and accurate comparison of labels against approved templates directly on the production line. The client aimed to minimize the risk of errors and costly production downtime.
Project Execution
The ALTEN Polska team developed the dedicated LabelScan application, utilizing Python and .NET technologies, with an AI engine based on OpenCV and scikit-image libraries. The system continuously compares production labels with the provided graphic template, detecting discrepancies such as barcode errors, typos, or missing graphic elements.
The project also included implementing technical infrastructure at the Kruszwica production site and integrating with a Data Center in Geneva. The system operates 24/7 across six inspection stations, offering functionalities such as automatic reporting of results and archiving data for up to one year. A key element of the implementation was the client’s involvement in defining system functionalities, ensuring alignment with real production needs.
Final Outcome
The LabelScan system enabled the automation of label verification with an accuracy rate of 99.9%. Production line operators can now identify errors in real-time, eliminating the need for manual checks and accelerating the production process. Additionally, the solution streamlined reporting and helped prevent costly production downtimes.
As emphasized by Przemysław Stępniak, Plant Manager at Bunge, the project was completed on time, with the highest quality standards, and with seamless collaboration between ALTEN Polska’s teams and the client’s stakeholders.
Summary
The LabelScan system implemented by ALTEN Polska is an innovative solution supporting automation and efficiency in FMCG production. By utilizing advanced algorithms and close collaboration with the client, the project met high expectations for quality and timeliness, delivering tangible business benefits to Bunge Polska.
