• Font Size:
  • A
  • A
  • A

Feature Articles

Current Machine Vision Activities in the Food Industry

by Nello Zuech, Contributing Editor - AIA

 

The food industry is in dire need of substituting capital equipment for labor as the often seasonal nature of the work leads to the use of migrant workers to fill these jobs. As has been well documented in the news, often these migrant workers are non citizens of the US, which poses challenges for companies. Furthermore, with the drive to increase the minimum wage, the cost of using people to perform labor-intensive tasks, such as inspection and grading, will increase dramatically – as much as 50%.

The food industry has a number of opportunities for machine vision and many companies have responded by developing application-specific solutions. This article reviews current trends of some of those solutions addressing applications at the growers, fresh pack and food processors. It is noted that each of these segments of the market has different inspection and grading requirements, but there are some general requirements shared by all:

  • Sort by size
  • Sort by shape
  • Sort by ripeness
  • Sort by color
  • Sort by grade or a by combination of size, shape, ripeness, color
  • Eliminate foreign material

Sorting by size is straightforward. The actual dimensions of the product are measured. In some cases size is a variable proportional to value. The larger the potatoes the more the consumer is willing to pay. Sorting by shape is also relatively straightforward with today’s machine vision technology. Any number of geometrics can be used; for example, in the case of elongated products, length-to-width ratio; other product dependent criteria include: area, feature geometry, circularity, curvatures, holes, surface area or volume, etc. Sorting by grade usually involves looking at the surface for blemishes of one type or another – surface blemishes, insect damage, punctures, hail damage, cuts and bruising. What constitutes a blemish is product specific. Sometimes it could actually be a surface condition reflecting a disease. In some cases the objective, in addition to product inspection, is the removal of foreign material and debris, such as dirt, sticks, pebbles, insects, small dead animals, etc. from the stream of product.

One distinct advantage machine vision-based inspection systems have is consistency. Once grading standards have been established the grading criteria will be consistently applied, unlike people who generally are more subjective in applying the criteria. More often than not people will grade to relative standards depending on the batch conditions currently under inspection.

Today the capability even exists to look beneath the surface for internal blemishes. The key is to successfully distinguish what constitutes an irregularity while permitting a wide range of appearance conditions that are perfectly acceptable, especially conditions like stems in fruits like apples. In some cases what needs to be detected is hidden within the product, a pit, for example, in the case of cherries or dried fruit like prunes.

What is interesting is to see that the suppliers of these application-specific machine vision systems are embracing the latest in the underlying technologies available for machine vision systems. Virtually all suppliers now offer systems based on digital line scan or area cameras (USB, FireWire, Camera Link or GigE Vision) and megapixel cameras. Those that offer systems specifically for grading often use color cameras and/or near-infrared-based cameras to detect subsurface conditions. Some even offer x-ray-based systems to characterize the product integrity based on internal conditions. Not all companies use commercially available cameras – many are based on proprietary designs optimized for product grading applications.

Many of the systems have embraced LED lighting arrangements, often with the specific color arrangement of LEDs product specific, in order to enhance conditions that are the basis of assessing grade. Generally these arrangements are based on proprietary designs.  Certainly the systems take full advantage of today’s computing technology frequently incorporating microprocessors, DSPs and FPGAs to offload and accelerate the computing requirements, especially where grading based on comprehensive properties is required. In some cases one finds multi-spectral processing capabilities where simultaneous analysis is being done on a visible image and a near-IR image or on the combination of the two or on two visible images based on different spectral responses. Some near-IR-based systems are designed to detect such internal characteristics as sweetness, dry matter, oil content, etc.

In addition to identifying defective conditions, some of these systems have the ability to measure the area coverage of the defect condition and/or count the number of defect conditions to set tolerances for allowable coverage and/or allowable number of defects. Some of these systems even have the ability to classify defects by properties. In some cases classification is based on neural networks.

Mechanically the systems often include arrangements of cameras to adequately cover a wide belt of products passing by the cameras. In some cases cameras are used to view the product from the front and back and essentially perform a three-dimensional analysis of the product. In other cases, where the shape of the product permits it to be rotated, product is rotated in front of a single camera or camera arrangement while analyzing several images of the same product.

Most systems have the capacity to handle large volumes of product often measured in terms of tons per hour. Some products lend themselves to being delivered nested in multiple lanes to increase the inspection throughput. Others, especially dry products like seeds, nuts, beans, etc. are delivered in channels, again frequently in multiple channels to increase the inspection throughput. Many wet and frozen products are delivered by belts with cameras mounted over the belts to capture scene images in the direction of travel. One of the biggest challenges of these systems is to optimize the process of eliminating unwanted product. Where products are delivered in discrete nested fashion, this is not a problem. But where products can be touching and removal from a belt or channel is required, there is always a reasonable probability that good product will be eliminated with a bad product.

Today these systems come with relatively simple man-machine interfaces often following a Microsoft Windows® icon format. Where color is critical to an application, color calibration tactics are incorporated into the system. Most systems include report screens reflecting reject rates, etc. Many systems can be adapted to more than one product – generally the products must be compatible with the mechanical delivery system.

To complement imaging-based machine vision systems, many companies also offer systems based on laser scanning approaches where specific wavelengths of the laser provide enhanced sorting capabilities. Some also combine weighing systems to include another parameter for grading. Sorts according to the weight to volume ratio of some products can be used to detect product that has been frost bitten or has other internal damage.

The following tables were assembled on a best effort basis based on a review of the websites of companies known to provide inspection systems for the food industry.

 

Comments:

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.

First Name: *
Last Name: *
Your Email: *
Your Comment:
Please check the box below and respond as instructed.

Search AIA:


Browse by Products:


Browse by Company Type: