For years, manufacturers perceived their product labels primarily as a channel of customer communication. They used the label to convince potential buyers of the quality of their product immediately before the purchase decision and conveyed important, sometimes regulated information on it - from the best-before date and allergen warning to practical preparation tips. Today, however, the label also plays a key role in terms of communication along the supply chain: without detailed labeling, there would be no inventory management, no product specification, no tracking and tracing - and product recalls would be an even bigger nightmare than they already are.

Given the importance of product labeling for consumer protection, labeling is strictly regulated in many areas. In many countries, for example, the labeling of product barcodes, batch numbers and best-before dates is now mandatory. In addition, even where there are no legal regulations, consumer pressure is often so high that the relevant information is still to be found on the packaging. Machine-readable, serialized labels, which have been mandatory in the pharmaceutical industry for prescription drugs since February 2019, are also increasingly being used in the food industry and by suppliers.

Incorrect declarations can be expensive

In this highly regulated environment, incorrect labeling harbours a high potential for damage: it is not only the handling of recalls that is associated with high costs. Because the food industry is producing faster and faster, thousands of faulty products can roll off the production line before an incorrect label is discovered. If the product then has to be reworked or even repackaged, this is associated with a reduction in productivity and immense financial costs. And the damage to a company's image in the event of such an incident is also much more dramatic today than it was a few years ago - simply because the internet is such an effective multiplier for negative feedback.

Optical inspection systems that automatically monitor the quality and correctness of the label inline are therefore becoming increasingly popular with more and more companies. In contrast to the previously common random inspections, modern systems verify the information on each individual product.

One of the typical but comparatively difficult tasks of optical inspection is the analysis of alphanumeric character strings. Plain text sequences of numbers and letters are used in many areas of the food industry, for example to indicate serial and batch numbers, best-before dates or serialization codes. Their great advantage is that, unlike barcodes, they can be read and checked by users without additional equipment. Alternatively, they can now also be inspected automatically as part of optical inspection. There are two different methods:

• Optical Character Verification (OCV) checks whether the applied code matches a stored reference code and is sufficiently legible. This makes it possible, for example, to verify whether the correct batch number has been applied to the packaging.
• With the more complex Optical Character Recognition (OCR ), there is no reference - instead, an unknown text must be captured and read. This process is used, for example, to read newly generated and printed serial numbers.

Optical Character Verification

Optical Character Verification (OCV) checks whether the applied code matches a stored reference code and is sufficiently legible.

© Mettler Toledo

With OCV, the inspection system captures the passing products with a camera mounted above the conveyor belt and analyzes the images using image processing software. This first determines where the character string to be analyzed is located on the image. The system then defines a search area around the text and - similar to a barcode reader - uses the contrast between the print and the background to isolate the individual characters. The inspection system then compares each individual character with the stored reference image of the label and evaluates the degree of conformity. If a character has been printed incorrectly or illegibly, this becomes obvious at the latest when it is compared with the template - and the evaluation is correspondingly poor. In this way, the quality and correctness of the label can be assessed quickly and reliably.

However, the process reaches its limits if no reference image is available for comparison - for example, if a unique serial number is generated and printed for each package. In this case, the inspection system does not know which character string to expect next. In such a scenario, an optical inspection system must therefore be able to independently decipher and correctly interpret the unknown characters of the serialization code - a typical area of application for OCR.

Optical Character Recognition

With Optical Character Recognition (OCR), each letter is compared with each individual character of the stored font and the character that comes closest to the one captured is determined at the end.

© Mettler Toledo

Before being used, an OCR must first be stored with the font that is used for printing the labels or packaging. The first steps in the inspection process are then similar to OCV: the system first captures an image of the product, defines the search area, identifies the character string and breaks it down into individual numbers and letters. The system then compares each of these characters with each individual character of the stored font - and finally determines the character that comes closest to the captured character. Once all the characters have been analyzed, the system compiles the complete character string from the results obtained.

However, the disadvantage of this process is that it only evaluates the results to a limited extent. The OCR tool assigns the best matching equivalent in the typeset to each character. This means that even barely legible printouts are converted into character strings. Due to the poor print quality, the labels are usually rejected anyway, simply because the supposedly best matching character is simply not correct. In many scenarios, an OCV is therefore carried out after the OCR to be on the safe side - which increases reliability but prolongs the inspection process.

An example from practice

Inspection systems such as the 'V2622 Flex-Lite' from Mettler Toledo automatically monitor the quality and correctness of the product declaration so that it meets the increasingly stringent compliance requirements.

© Mettler Toledo

Many manufacturers in the beverage industry also rely on optical inspection systems for labeling checks. A typical integration point is between the filling line and the checkweigher. Once the beverage carton has passed through the filling line, it passes the smart cameras of the optical inspection system. Within fractions of a second, these capture razor-sharp images of the product and compare them with the stored reference images using OCV. When selecting smart cameras, manufacturers can choose from various models with different resolutions, lenses and lighting types for black-and-white and color readings.

Devices that offer camera options with a liquid lens, such as the 'V2622 Flex-Lite' modular optical inspection system from Mettler Toledo, have a special feature: While conventional models with a fixed focus setting have to be readjusted every time the product is changed, the liquid lens of this camera changes its focus within fractions of a second without any moving parts. This allows product features to be checked at different levels during a single inspection step. This makes it possible to vary the focal length of the lens and thus capture several images in succession at different focal planes within a very short time. For example, when inspecting gable-top milk drink cartons, the liquid lens can check the net filling quantity on the side and the best-before date printed on the upper vertical flap in just one inspection step and without readjusting the cameras. Even when items are changed - for example from 1 to 1.5 liter cartons, where the product size and therefore the reading distance changes - the camera automatically readjusts the focus. If the inspection system detects an error in the declaration - such as an incorrect best-before date - the beverage carton is reliably removed from the line by an ejection mechanism (such as a pusher).

Author:
Reinhold van Ackeren is Head of Marketing & Product Management at Mettler Toledo CI-Vision in Zwingenberg.

Stricter labeling requirements

Since the end of 2014, EU Regulation 1169/2011 - better known in Germany as the Food Information Regulation or FIR - has regulated the mandatory labeling of food. In addition to the minimum font size, it stipulates the labeling with the respective product name, a list of ingredients, detailed origin information and allergen information, as well as a range of other product-dependent information. This includes, for example, the freezing date for frozen products. The labeling obligation will soon be tightened again: from 13 December 2016, labeling with a uniform, detailed nutritional table will be mandatory throughout the EU.