Man v Machine: A vision for product traceability

In the war of man v machine, the realm of code reading is turning into a hard-fought battleground as technology advances on the capabilities of the human brain.  Highly-complex visual data can now be included in high-speed machine-readable formats, such as 1D barcodes, 2D datamatrix codes and QR codes. Meanwhile, humans still rely, on old-fashioned alphanumeric letters and numbers to understand and convey information.

However, the brain can ‘decode’ a wide variation in design and quality of lettering, using intuition about missing information to fill in gaps, and applying ‘common sense’ to verify the information.

 

True, barcodes, data matrix and QR codes are designed with redundancy in mind, so that a certain amount of damage can be tolerated in a production environment. The codes are still in the binary form i.e. no print or some print, whether that print is a line, dot or a square.

 

Letters, numbers and symbols are a different story.  Humans can recognise lettering shapes despite distortions; because we can also look at context and meaning e.g. ‘th0ught’ is recognisable.  Automated systems need to use camera-based technologies to detect and inspect shapes and symbols by comparing them to a pre-determined shape or pattern.

On-pack data

Human readable on-pack data, such as batch, lot numbers, best before or expiry dates are critical for products such as food, pharmaceutical, medical devices and cosmetics. These numbers, letters, symbols and codes need to be verified, be of a readable quality and sometimes match the barcode to ensure the right data is associated with the right pack. If required, manufacturers, the supply chain and retailers must all be able to cross check the machine-read coding with the alphanumerics for read-quality and traceability.

 

Using machine vision-based technology promises new opportunities to read, verify, check the quality of, and match alphanumeric labelling information. It is one of the ways to save punitive fines and comply with the requirements of retailers and the law.

 

OCR and OCV systems

OCV systems and OCR systems have been available for some time to offer automated solutions. OCR systems work by comparing a library of taught models to what is printed. OCV is a direct comparison between what should be there and what is printed. The output from an OCR system is the alpha numeric string that has been read such as a use by date. The output from an OCV system is usually a pass or fail signal which is an indication of quality and legibility.

Historically, these technologies have had limitations. If you’ve watched a check-out assistant struggling to get the till scanner to read simple barcode lines from a wrinkly label on a pack, imagine how difficult it could be to scan distorted letters. I believe we are still far short of achieving the ideal of being able to deliver perfect alphanumeric print to identify a product, part or pack every time under high speed printing in industrial conditions or part marking by laser or engraving stamp, for example.

Extensive research

To make more of OCR and OCV technology depends on improving the power of the algorithm used by the scanner software. As a consequence, SICK spent several years researching the best decoding algorithms before launching its own system the LECTOR 620 OCR, which combines 2D and barcode reading, OCR and OCV in one device.

The new technology makes code reading, quality-checking and matching, faster, simpler and more reliable. It enables plain text letters, symbols and numbers to be detected at a distance of between 30 and 300mm both while stationary and at speeds of up to 4.0m/s. A range of standard fonts are supported, as well as barcodes and data matrix codes, ensuring both legibility and placement for essential quality control. The ability to train different fonts is also possible as production is running.

Field tests

To illustrate what is achievable, the SICK development team subjected the LECTOR 620 OCR to extended field testing. There are five general classes of alphanumeric printing which were tested:

•  Lettering in a fixed position

•  Lettering in a free position

•  Lettering in a fixed position enclosed within a frame

•  Lettering printed adjacent to coding

•  Lettering printed adjacent to a logo or other pattern.

A new locator function used in the Lector 620 was key to enabling the reader to identify and read the object despite variations in its presentation.

In general the field tests showed that all OCR features of the LECTOR 620 OCR worked very well. But they also demonstrated that some users, especially in the food industry, were not used to having to produce print text and fonts of such a high quality and reproducibility. In the pharmaceutical industry legal restrictions meant that the need for perfect printed data was better established.

It is important, therefore, to recognise ‘best practice’ to ensure success. Manufacturers need to consider the set up of their printing and labelling carefully to achieve optimum results. The quality of the printing and marking techniques may need to be upgraded and the contrast or flatness of surface being printed improved.  

Optimising printing and marking processes is of paramount importance. The choice of fonts, character size, use of upper or lower case letters and special characters and background contrast, for example, also need careful consideration. As with all code reading, a sufficient ‘quiet zone’ (i.e. one with no print in it) between the font and the edge of printing area is essential.

Best practice

It’s important to work closely with a consultant or supplier to ensure the best conditions for optimum performance and evaluate the best system solution for each application. That said, opportunities are now available to install a highly functional reading, quality-checking and code-matching solution. Manufacturers may need to commit time and money to optimising systems, but the rewards in ensuring good customer relations and avoiding fines and returned products could more than justify the investment.