The best things in life may be free, but machine-vision systems must be properly specified-and that’s not inexpensive.
After Christmas is a really good time to buy tinned cranberry jelly, especially if you need to save your pennies. This year was no different. As I pushed my trolley around the local supermarket I noticed a huge pile of tins located in the bargain section of the store. Then, I noticed why the bitter berries were so inexpensive.
The manufacturer had placed the labels upside down on the tins. Being totally unacceptable to the American consumer, the store had decided to “fire sale” the complete batch at a cost of $0.35 per can. This was the bargain of the century. Obviously, the company will need to deploy a machine-vision system to ensure the heinous event does not occur again!
But where can manufacturing engineers faced with improving the quality of their products seek reliable information about machine vision? Although many are familiar with the process-control and monitoring systems that have been available for decades, very few, it seems, understand the potential and limitations of machine-vision technology.
To help these folks, many magazines and Web sites have set up bulletin boards and feedback sections where engineers can pose questions and, hopefully, receive competent answers for machine-vision manufacturers and system integrators alike. Often, however, the questions posed on these Web sites are overly simplistic and do not provide manufacturers or system developers with an idea of how to approach a possible solution to the problem.
One question recently submitted to the Vision Systems Design Web forum asked whether there was possibly a machine-vision system available that could properly inspect the color and geometric properties of cookies (or biscuits, as we call them in England). Of course, without any further information, the answer to this and other similar questions is “maybe.” For without knowing the exact size of the objects, their color, the speed of production, how they are oriented, whether they are evenly spaced, the type of lighting in the plant, the existing process control systems, and other pertinent factors, the question alone simply cannot be answered.
It is apparent that many still believe that companies exist that can deliver fully configured systems that act like programmable logic controllers and that can be easily installed to operate in a matter of weeks. Such systems, once deployed, will, of course, seamlessly detect every faulty product and run 24/7, 365 days per year. Unfortunately, nothing could be further from the truth.
Even before manufacturing personnel embark on specifying such a system, they must carefully consider what it is expected to accomplish. For this reason, Nello Zuech at Vision Systems International (Yardley, PA, USA) has developed what he calls a Machine Vision Application Requirements Check List that is freely available at http://www.imagelabs.com/vsi/mvaques.html. Although this may take a lot longer to fill in than your federal income-tax form, the result will be a clear account of what you require. And, once completed, it will be far easier for manufacturers and system integrators to understand whether they can provide you with a possible solution.
Although the components may be relatively inexpensive, nearly every one will require tailoring to meet specific goals. Image-processing algorithms may need to be configured, user interfaces developed, and pass/fail database data integrated to provide the feedback required to both monitor the parts and provide data about the manufacturing process. Therefore, many machine-vision systems must be custom-built and may cost more than an order of magnitude of the cost of the parts used.
Once the systems are deployed, their benefits can easily justify their expense in terms of both increasing product quality and ensuring a consistent manufacturing process. While those looking for bargains will be disappointed, those prepared to spend the necessary time to develop complete specifications will be amply rewarded.
Andy Wilson
editor
[email protected]
