Customized vision system for label quality control

Article Author KSM Vision Marketing

At KSM Vision, we have just embarked on a new implementation for a client in the food industry. For one of the most technologically advanced and leading poultry producers in Poland, we are developing a customized off-line vision system to support manual label inspection.

An 'offline' label quality control system - why?

The offered system for quality control of labels on pouches and cartons will be set up outside the production line area to allow the operator to verify the label. The system will support automation of manual inspection of printed labels on transparent films as well as opaque stickers on cartons and pouches.

Controlling the quality of prints on such diverse labels and packaging with a single system will be possible thanks to:

  • the use of two lighting systems simultaneously, making it possible to inspect both transparent films and opaque stickers,
  • for ‘pouche’-type packaging, a special system dedicated to the project will be developed to allow manual tensioning of the bags to enable correct reading of the characters on the label. A special pouch inspection procedure will be developed as part of the system implementation,
  • the ‘pattern matching’ function, which will verify the similarity of the entire inscription (product names can be written in fonts created by graphic designers, which makes them sometimes unreadable for OCR algorithms),
  • analysis of read data by KSM Vision’s proprietary software based on deep neural networks offering much more than machine learning (deep learning AI).

Off-line print quality control - how does it work?

After inserting the label under the image acquisition system, the operator will select the appropriate product (label) and enter the delivery date and trigger image acquisition.

The label data will be read, such as:

  1. Product name,
  2. Expiration date,
  3. Series and lot number,
  4. Barcode / characters under the barcode,
  5. ‘Pattern matching’ for selected areas.

The data read from the label will be compared, with the data entered by the production assistant (the day of the year will be decoded, in the case of dates, there will be a calculation and check if there is a corresponding difference between the frying date, the date of shipment and the allowable date for consumption). The operator will be able to choose whether he has finished scanning or is scanning another label of the same product (carton, pallet). The data from the unit label will then be compared with the data from subsequent labels. A picture of the label, the data read, the operator’s ID will be stored in the database, and the data can be ripped and archived.

Dedicated vision system for label inspection - for whom?

The project is being carried out for one of the most technologically advanced and leading poultry producers in Poland. The brand is ahead of the trends in robotization and production automation, so the implementation of a label scanner that automates the reading of product names, dates, batch numbers and barcodes from bags, labels and cartons, along with comparing the read data with predefined requirements, is an important element in optimizing the production process.

Our R&D on the solutions with the use of neural network potential are backed by our renowned partners: