Machine Vision Touches Growing Number of Everyday Products

Machine vision touches growing number of everyday products 

 From diapers, false teeth and pain relievers to sports equipment, roofing material, beer, liquor and cookies, a growing number of products we encounter in our daily lives rely on machine vision to expedite the production process and improve quality.

The trend springs from a growing need to assure top quality and accuracy in products and their packages, of course. But it is also driven by a growing need to reduce total costs of product manufacturing and packaging by minimizing human labor and replacing it with tireless automatic product inspection and, also, for guidance in the picking, placing and sorting of components, sometimes accompanied by robotic actions. 'In the process, operational speed, detection levels, quality standards and product monitoring can be greatly increased as well,' notes Bill Whitmer, senior sales engineer for Integral Vision, Farmington Hills, MI.

Significant declines in vision system prices have encouraged adoption, too. Integral Vision, for example, has cameras with integral algorithms, so-called 'vision sensors,' which sell for $2,000 to $3,000.

On the other hand, more complex systems will require a relatively large capital investment ranging up to a quarter million dollars. To inspect the entire surface of wide rolls of flexible material like decorated fabrics 'may require as many as 20 to 30 cameras integrated into a complex controller-processor system,' asserts Vic Wintriss, president of Wintriss Engineering Corp., San Diego, CA, a supplier and integrator of machine vision systems.

Wintriss, for example, has supplied inspection systems for a manufacturer of roofing materials. The 14-foot-wide web moves at 1,200 feet per minute and is subsequently slit into narrower increments for consumer and commercial users. This operation has six line-scan cameras on the manufacturing line, which run progressive scans at 70,000 scans (400,000,000 pixels) per second. How is this kind of high-speed processing monster possible?

'Our answer,' says Wintriss, 'is one of our special PC drivers with a powerful circuit containing 64 megabytes of Ethernet capability built into a hub that holds all of the so-called 'smart cameras,' made for this usage by ZX Works, Wind River Systems, Inc. of Alameda, CA. All of the processing is done in the cameras with data integrated into one output.' The systems used by Wintriss for the above application can capture and process a 1.3-million pixel image before it leaves the camera. A modification of the camera is said to differentiate between colors. For exceptional color differentiation, it can be equipped with an optional spectrophotometer, which captures hyperspectral color information. Such a unit is now necessary for many medical products applications as well, particularly for determining with 100% confidence that all tablets in a stream running from manufacturing to final packaging are indeed the same product.

Some small product applications also can be costly, we are reminded by John J. Roberts, vice president of AccuSentry, Marietta, GA, which provides systems for the pharmaceutical field. These applications often require vision systems in the mid- to high-cost range because of the sophisticated camera sensing required and the high speeds needed to process the massive amounts of data generated. At a low end of about $30,000, this firm and others like it, offer systems for traditional package inspections such as lot code and date code verification, safety seal and label inspection. But the pharmaceutical field, which has basically already overcome the problem of spotting cracked and split tablets, is faced with increased demands from the U.S. Food & Drug Administration (FDA), Washington, DC, to address more complex jobs such as verification of tablet dimensions and the inspection of lenses and tempered glass. Also in demand is 100% inspection and verification of copy on enclosed literature, which, today, can run to increasingly long documents.

As mentioned above, a particularly demanding job in the pharmaceutical field is the prevention of product mix-ups. A seemingly simple solution, one would think, is to give each tablet its own distinctive color so stray tablets could be easily recognized and eliminated. But colors are chosen for both patient and marketing reasons. Furthermore, even a single color can vary in hue. That has been the rub in creating a vision system with sufficient sophistication to identify different colors, slight variations in the hue of a single color or surface finish differences, ranging from glossy to matte, which can change the appearance of a tablet's color. Under pressure from FDA, the pharmaceutical industry is vigorously seeking a high-speed solution. A camera such as one previously mentioned may be an answer.

Another possibility is the TIPS HD-CBP (Hi-Density Color Blister Pack) system from Systech, Cranbury, NJ. It features a digital color camera array and is mounted on blister packaging machines to check that the tablets present are the proper size, shape and color, that none are missing, damaged or in the wrong location and that no foreign objects are present. Setup is quickly accomplished by presenting an empty blister pack and a properly filled one to the system.

With advances in machine vision technology, FDA has developed a preference for 100% inspection versus statistical sampling plans because it is the only way total perfection can be assured.

An example of what can be done is the self-contained Auto Image ID system from RVSI Acuity CiMatrix, Canton, MA. It checks two-dimensional Data Matrix codes on glass and pharmaceutical vials. The camera with integral lighting, lens and microprocessor, inspects coded vials to eliminate unreadable codes and any incorrect vials as bulk-stored containers are brought out for final filling and sealing. The system, 'not only sharpens the inspection process, but saves quite a bit of cost by eliminating human inspectors,' says John Agapakis, general manager of Acuity. A laser alignment function also speeds setup.

Yet another automatic inspection system, this one from M.W. Technologies, Inc., Elmwood Park, NJ, was invented seven years ago to address FDA's desire for closer inspection of tablets. As with most vision systems, it has been constantly updated with the latest model introduced at the recent Interphex show in Philadelphia. M.W. builds the conveyorized materials handling system and then adds any vision system desired by the customer.

Two recent systems incorporate machine vision systems from Cognex Corp., Natick, MA, and RVSI Acuity, respectively. 'At the beginning, we didn't get much of a response from industry,' states company Sales and Marketing Manager Mike Dragotta, 'but with the renewed emphasis by FDA on complete surety and, also, an equal demand by German and Japanese companies for 100% surety in the expensive high-tech drugs they are now importing in growing amounts, we are immersed, as are many vision companies, in research to outwit the color identification problems in tablets.' Starting big, M.W. is working on a contract for identifying yellow tablets. It is investigating both color and gray-scale approaches to the problem, one of the most intensive, since this color and its hues are among the most difficult to detect.

For small or medium size products, especially those with irregular contours, machine vision is being integrated with robotic units, which can move the camera to expose any and all surfaces for vision inspection or provide positioning data for assembly. Although most commonly used in the automotive industry, currently (see
www.machinevisiononline.org, 'Smarter Machine Vision Systems Proliferate in Automotive Industry'), usage is expanding in the consumer products sector.

AccuSentry is one company which integrates robots to mobilize some of the vision applications. As an example, the company recently installed a sophisticated vision unit that controls two 45-kilogram (99-pound) payload robots that supply and remove product from a complete quality-control work cell. 'Robots can sometimes be very expensive,' reports AccuSentry's Roberts. 'Therefore, many users look for a 'bargain' in the accompanying vision system(s). But in many cases (perhaps the larger number) there are a lot of 'hidden costs' in just the integration and engineering of high-tech vision systems. When a robot is added, too, costs are bound to grow, sometimes dramatically,' he concludes.

Another vision firm involved in both usual and unusual applications is Intelligent Automation Systems, Inc. (IA), Cambridge, MA. In addition to possessing a conventional two-dimensional technology, it has a three-dimensional sensor. Called the 4DI imager, it has pushed the capability of noncontact metrology to enable inspection within increasingly small tolerances. A nontraditional application is human-form measurement. 'We developed a system for a sporting goods manufacturer that measures the dimensions of a human hand for the creation of custom golf gloves. Still another creates an image from a wax mold with which to enable manufacture of a custom-fitted hearing device,' states Sarah Cremer, associate director of Marketing. This system also is used to inspect components, dies and package assemblies for the telecommunication device industry. Still another triumph for this firm is development of a system to measure both the thickness and volumetric measurement of a chocolate chip cookie.

Another human-form application is the determination of an exactly matching shade for a replacement tooth. The old manual comparison of living teeth to dental industry charts of available hues was less than satisfactory. As usual, manual inspection has drawbacks, particularly when it requires looking into a dark, wet 'cave' and comparing its contents to printed images of available replacement teeth and their hues.

So, a better way was offered to dentists at this year's meeting of the American Dental Association, Chicago, IL. X-Rite Inc., Grandville, MI, a manufacturer of color instrumentation, has created a vision camera probe equipped with a version of its spectrophotometer that can be slipped into a person's mouth to grab an image of one or more teeth. The images are stored and compared with the computerized images of available replacements. 'The technology matches the replacement very closely to the original in color, hue and value (lightness and darkness) and chroma (color saturation),' says X-Rite Engineer John Overbeck.

Probably the most common use of machine vision checks for packaging defects, including confirming date and lot codes and label presence and accuracy. 'As production speeds increase, human operators can't keep up,' explains Pierre Turcotte, sales manager at Optel Vision, Quebec, QE. Optel, for example, has systems in place to check products ranging from pain relievers and vitamins to beer and cosmetics. Characteristic are two brewery installations. The first ensures that labels and caps are the correct brand for the run, while the second looks for cap problems such as a missing tamper-evident band. In other applications, a distiller uses an Optel system to make sure labels are not torn or otherwise damaged, while a snack maker checks pouch seal integrity and a confectioner verifies product thickness.

As a result, these consumer products companies can deliver the highest quality products to their customers today and every day. This not only encourages sales, but also promotes brand loyalty, which can be damaged by the purchase of a single off-spec item.