- This article is filed under:
- » View All
Search for ‘‘Perfectly Safe’‘ Products Pushes Machine Vision into Food Industry
by Winn Hardin, Contributing Editor - AIA Posted 03/21/2005
A trip to the local grocery store makes a great argument against the idea that perfection is unattainable. The piles of fruits and vegetables on display look as if they were made by cookie cutters, each the same in color, ripeness, fullness and shape. These days, you have to look hard to find a green pepper with a bruise.
Hundreds or even thousands of miles away at a food processing center, a growing number of color machine vision systems make sure that the fruit and vegetables that reach store shelves meet stringent visual criteria. Machine vision also is reaching deeper into the back-end of food processing, thanks to pending food labeling regulations from the U.S. Food and Drug Administration (FDA) that will impose documentation requirements on the food industry similar to the pharmaceutical industry and FDA regulation 21 CFR Part 11.
Food is color
Dunkley International (Kalamazoo, Michigan) isn't in Michigan just because it's the automobile capital of the world…it's also one of the world's largest producers of cherries and blueberries. Dunkley's first sorting machine built in 1885 used a series of leather belts to sort fruit and vegetables by size. Later, Dunkley was one of the first commercial adopters of machine vision systems, installing its first vision systems in the early 1970s.
‘‘The old technology was basically grayscale cameras, but it wasn't a very good solution because it could only collect generalized information. When [the industry] came out with color line-scan cameras, we moved in that direction because food is normally running on conveyors at 600 to 700 feet per minute, and you need the ability to image fast moving objects,’‘ explained Ernest Kenneway, Dunkley's general manager. ‘‘We are able to use PC hosts with proprietary software and algorithms to get the speed out of this data intensive process.’‘
Food inspection systems typically use color cameras to determine ripeness of a fruit based on color, and identify damaged fruit as well as pebbles or other contaminants. ‘‘With the current technology, you can do any food, french fries, peas, it's whatever you need,’‘ Kenneway said.
In addition to quality, Kenneway added that traceability, product tracking and liability concerns are driving the food industry in the same ways that it's driving durable goods such as automobiles and pharmaceuticals, for example. ‘‘People are doing more to isolate lots for recall purposes if there's a problem. The industry is moving towards individual product identification so if there's a problem, they can isolate it; know when it happened and how many were manufactured during that time frame. For more and more people, traceability is becoming critical to operations,’‘ Kenneway said.
The FDA's ‘‘Sec. 555.250 Statement of Policy for Labeling and Preventing Cross-contact of Common Food Allergens’‘ released in 2001 is the main reason why food processors are inserting machine vision into more nodes along the production line. According to Amy Simonne, assistant professor at the University of Florida (Gainesville, Florida), experts agree that food allergies in developed countries are becoming more common. Food allergies affect 2 - 2.5% of adults and 6 - 8% of children in the U.S. and result in 100 - 175 deaths each year. By 2002, more than 160 foods were associated with allergic reactions.
In response, the FDA's Sec. 555.250 requires any food that comes into contact with the eight most common allergens, which are estimated to compose 90% of allergic reactions in the U.S. each year, be clearly labeled. Those allergens include: peanuts, soybeans, milk, eggs, fish crustacea (shell fish), tree nuts and wheat. The FDA is also considering increasing these labeling requirements to include allergenic ingredients that may come in spices, flavorings or colorings (see link at end of article for Sec. 555.250). Starting on January 1, 2006, Sec. 555.250 will move from suggested practice to federal regulation.
Inspection guides for FDA field workers require food manufacturers that produce allergenic products to verify and document that the right product is placed in the right package and then in the correct box. Better than any other verification method, images of labels provide proof-positive verification and protection in the case of liability claims.
According to Kris Bierbaum, packaging industry manager at Cognex Corp. (Natick, Massachusetts), allergen concerns are driving machine vision into several back-end food processing applications, including: cartoning, bright stock labeling, product packaging and label inspection applications, any of which may include any number of machine vision functions, such as bar code reading or optical character verification (OCV). Front-end vision applications include sorting, assembly verification and defect detection in addition to traceability functions.
Traceability is made even more important in the food industry because products may be sorted in one location, canned in another facility, and labeled, cartoned and palletized in another, explained Steve Cruickshank, principle product marketing manager of Cognex PC Vision Products. ‘‘This is especially true of generic, store brands. Typically, products are marked with dot matrix code that contains lot codes, dates, and contents and then that code has to be checked against the final label,’‘ Cruickshank said. ‘‘Pharmaceutical companies have similar needs to verify text strings on the packaging. Although the pharmaceutical and food industries have different technical challenges and different printing processes, the application is the same …verifying a text string.’‘
Both food and pharmaceutical production lines are among the fastest production lines in industry, moving several hundred feet per minute and faster. This, combined with the need for simplified, robust vision systems for new markets and end-users that aren’t experts in machine vision, has pushed Cognex to develop products geared specifically towards the food industry, such as ProofRead, which combines OCV with the company’s patent pending PatFlex technology. ProofRead accurately verifies the characters produced by a wide range of printers, even when the characters are extremely distorted.
According to Cruickshank, ProofRead's OCV tool defines the text string pattern based on font definitions taken directly from the printer, rather than the traditional process of showing the system images of printed characters. This makes it easier for operators with limited vision experience to train the system. By combining high speed OCV with PatFlex pattern matching algorithms, ProofRead can verify text strings even when the strings become distorted due to variations in line speed or printing process. ProofRead allows the integrator or operator to optimize the trade-off between line speed and allowable character distortion, all while maintaining the original font definition provided by the printer vendor.
Cruickshank continued, ‘‘Since many of these applications require reading 1D barcodes and 2D symbols in addition to text verification, ProofRead provides robust ID tools for reading a wide variety of bar codes and 2D symbols.’‘
As Dunkley's and Cognex's experiences show, machine vision is not just gaining market share in the food industry because of improved technology and greater education and acceptance among end users, but also because machine vision offers the only solution to provide the front end quality assurance and back-end safety assurance that today's food processors need. Look for good things from this market segment in 2005 as manufactures ready themselves for new federal regulations.
The CPG entitled ‘‘Statement of Policy for Labeling and Preventing Cross-contact of Common Food Allergens’‘ is available on the Internet at http://www.fda.gov/ora/compliance_ref/cpg/cpgfod/cpg555-250.htm. Copies of this CPG can also be obtained by faxing requests to 301-827-0482.
There are currently no comments for this article.
Leave a Comment:
All fields are required, but only your name and comment will be visible (email addresses are kept confidential). Comments are moderated and will not appear immediately. Please no link dropping, no keywords or domains as names; do not spam, and please do not advertise.