Food Safety Magazine

Signature Series | November 26, 2014

Preventing Undeclared Food Allergens Using Vision Systems

By Mark Schmid

Preventing Undeclared Food Allergens Using Vision Systems

According to Food Allergy Research & Education, as many as 15 million people are afflicted with food allergies—and that number is rising. In a 2013 study by the Centers for Disease Control and Prevention, food allergies among children were reported to have increased approximately 50 percent between 1997 and 2011.

An allergen, according to the National Institute of Health, is a substance that can cause an allergic reaction. While there are more than 200 known foods that can cause allergic reactions, eight allergens are reputed to cause more than 90 percent of allergic reactions in the United States:

  • Milk
  • Eggs
  • Fish
  • Shellfish
  • Tree nuts
  • Peanuts
  • Wheat
  • Soybeans

Allergen Control
According to the Stericycle ExpertRECALL™ Index, undeclared allergens—allergens either in a product, but not declared on the label or products contaminated by an allergen—caused 44 percent of U.S. Food and Drug Administration (FDA) recalls during the first quarter of 2014. Undeclared allergens were also responsible for 65 percent of recalls during the second quarter of 2013.

Undeclared allergens occur when a product is made with an allergen but the allergen is not listed on the label or when a product is not made with the allergen, but has become contaminated with the allergen through contact with another product.

Stericycle’s ExpertRECALL Index also revealed that the majority of U.S. Department of Agriculture- and FDA-recalled food products were categorized as Class I recalls, indicating that there is a high probability that consuming the recalled product could cause serious adverse health consequences or even death.

Recall Costs
Calculating the total cost of a recall is difficult. Expenses directly associated with the recall, such as product reacquisition, are relatively easy to determine. However, abstract costs like damage to a brand are much more difficult to measure. Conservative estimates factoring in cost of the product, marketing, logistics, administration, and corporate image begin at $100,000.

Regulating Recalls
The 2011 Food Safety Modernization Act shifts federal regulators’ focus from responding to allergen contamination to preventing it, thus helping to safeguard the U.S. food supply. The act gives the FDA the authority to shut down manufacturers that do not comply with regulations. Preventative controls can be applied to process controls, food allergen controls, sanitation controls and a recall plan. At least one source concludes that an effective allergen control program should include the following elements:

  • Verifying supplier ingredients do not have cross-contact issues and that the suppliers are adequately labeling products with allergens
  • Storing products containing an allergen over other products containing the same allergen to reduce the likelihood of allergen contamination in a spill
  • Schedule manufacturing such that processes involving little or no allergens are run first
  • Knowing which product formulas involve allergens
  • Sanitizing shared equipment, utensils and even employee uniforms between processes that involve allergens
  • Verifying labels to ensure that allergens are declared and that they appear in parentheses, for example, lethicin (soy), flour (wheat) and whey (milk)

Current procedures for line changes involve checklists and paperwork, but still require human verification that the changes have been made. Since packaging differences can be subtle, label verification is of the utmost importance in these programs. If every other step is performed correctly in the process, but the package labeling is wrong, consumers will be at significant risk.

Packaging and Labeling Errors
Preventing packaging and labeling errors from occurring can significantly minimize the risk to consumers posed by undeclared allergens. Producers and packagers must ensure that:

  • Product lids and tubs correspond to the SKU
  • Front and back labels are correct for each SKU
  • Roll stock relative to labels, pouch material and shrink sleeve is selected at startup and spliced according to the appropriate SKU
  • Primary package is placed in the correct secondary package
  • Ingredient packets are correct relative to the product

By identifying common causes of packaging and labeling errors, companies can develop better strategies to prevent mistakes from ever occurring. Some common causes include:

  • Printing mistakes
  • Mislabeled roll stock and packaging
  • Changes in product recipe not identified in the list of ingredients
  • Obsolete packaging materials being used
  • Human error

Verifying the Package Using Vision Sensors
The advent of 2-D barcodes has allowed for more information to be stored in a barcode. They require a relatively small amount of space and will not detract from the shelf appeal of a product. Ultraviolet (UV) codes are readable under UV illumination and completely invisible to consumers. These barcode types can be employed to verify that packaging material, including label, lid, tub, primary and secondary packaging, and inserts are correct for a given SKU.

Machine vision has become a central component in label verification throughout manufacturing and automation environments. With newer and more sophisticated features, vision sensors are designed to make communication, control and adaptation easier than ever before while remaining simple to operate and eliminating the possibility of human error. Vision sensors can perform numerous and diverse inspections simultaneously to accommodate high-speed assembly lines.

Easy Set Up
Vision sensors like the iVu series from Banner Engineering are a cost-effective alternative to more elaborate machine vision systems and are capable of performing important inspections at high rates of speed. Available with either an integrated or remote touch screen, users can setup, configure, monitor and modify inspection parameters on the factory floor without the use of a PC.

A vision sensor can be configured to inspect for the presence or absence of specific words associated with an allergen such as “cheese” or “peanut.” It will compare the product label to the stored image in the camera and accept or reject the label based on how well it matches. This can be a label verification method on a part of the package that has no barcode and enables the company to verify that a label corresponds correctly to the current production run.

iVu series sensors are capable of 1-D and 2-D barcode inspections and have many available ring light options, including ultraviolet for invisible codes, to facilitate inspections. It has an Ethernet IP option for easy connectivity and the new second generation version is capable of faster line speeds than before.

Most food companies don’t have the capacity to dedicate lines for product containing allergens. Therefore, it is necessary to verify that a line change was complete all the way to the packaging material.

While Banner’s iVu is capable of recognizing the patterns associated with a specific word, more sophisticated sensors, like Banner’s PresencePLUS® P4 OMNI feature optical character recognition and optical character verification. This allows the sensor to verify the text of the ingredients section on a product label. These cameras can transmit decoded American Standard Code for Information Exchange text strings using RS-232, EtherNet/IP or Modbus TCP/IP industrial protocols. The cameras can also simply match a string of text from a library of learned characters. They do require a PC for initial setup and configuration, but are capable of performing more complex inspections. They are available in corrosion resistant or stainless steel IP68 rated housings.

Looking Ahead
By going beyond simple operator manual inspection, vision sensors can automate the process of packaging verification. This can reduce the number of costly recalls associated with undisclosed allergens.

Mark Schmid is a business development manager with Banner Engineering.