The Importance of Expiration Date Inspection in Food Processing

Food products are directly consumed by customers, so regulations and legal standards for quality and safety are strict. If issues related to expiration dates arise, companies may face not only legal liability but also significant damage to brand trust. Therefore, thorough quality inspection in food manufacturing processes is essential.

Recently, as consumer reports have increased due to missing or incorrect expiration date labeling, the importance of expiration date inspection in the food industry has grown even further. In particular, for fresh products such as tofu, milk, and eggs, shorter shelf life leads to more frequent consumer verification, making expiration date labeling a critical factor.

Inspection of Expiration Date Printing Defects on Tofu Packaging

Company D, a tofu manufacturer, handles the entire production process from tofu forming to quality inspection.

Company D uses a dot printing method to mark expiration dates. The printed numbers are evaluated using a rule-based inspection method based on the number of dots. For example, ‘9 dots’ are defined as the number 2, and ‘8 dots’ as the number 7. Previously, a vision sensor was used to inspect expiration dates, but this method could only verify the presence or absence of printing.

Increased Scrap Rate Due to Misreading Errors

Since dot printing consists of very small dots, issues such as dot spreading, loss, or misalignment frequently occur. This caused critical problems in Company D’s production process.

For example, even if just one dot is missing from an ‘8’, it could be misrecognized as a ‘2’, resulting in normal products being incorrectly classified as expired and discarded. As a result, products that were still safe for consumption were frequently scrapped due to simple printing defects, significantly reducing productivity.

Solving the Problem with Multi Deep Learning Inspection Models

To resolve misrecognition issues caused by printing defects, Company D introduced Neurolcle’s auto deep learning vision inspection software. As a result, the problem was quickly resolved by replacing the existing rule-based inspection method with software alone.

Using Neuro-T, Neurolcle’s deep learning vision inspection software, Company D developed a Classification model to detect printing defects and an OCR model to recognize expiration date digits, creating a multi deep learning model system. Neuro-T provides a total of nine models, including Classification for defect detection, Segmentation for pixel-level defect recognition, and OCR for text recognition. These models can be flexibly combined depending on the required output and inspection criteria.

At Company D, a Classification model was first used to filter out severely defective products where the expiration date shape could not be recognized.

Next, an OCR model analyzed the expiration date based on shape rather than dot count, enabling accurate digit recognition. This allowed even partially missing or blurred dots to be correctly interpreted, reducing defect classification errors.

17% Increase in Expiration Date Defect Detection Rate

After adopting Neurolcle’s deep learning vision inspection software, there were no longer cases of products being shipped with incorrectly labeled expiration dates that resulted in customer complaints. Previously, products were often discarded due to incorrect digit recognition based on dot counts.

Now, the deep learning model accurately recognizes the original digits even when some dots are missing, reducing unnecessary scrap caused by simple printing defects and significantly improving productivity in the tofu manufacturing process.

Neurolcle has also implemented deep learning vision inspection solutions across various areas in the food industry, including meat processing quality inspection, sealing inspection of packaged foods, and noodle quality inspection. Recently, inquiries about implementation in food production processes have been increasing, highlighting the growing importance of quality inspection in the food industry.

If you would like to learn more about Neurolcle’s deep learning vision inspection use cases or discuss implementation, please click the button below to submit an inquiry, and we will respond promptly.