Machine vision has evolved rapidly to meet the needs of healthcare packagers.
When FDA's Tony Lord published a report on pharmaceutical labeling errors more than 10 years ago, the industry entered a new era. Based on Lord's findings, FDA issued proposed changes to the current good manufacturing practices that encouraged the use of machine vision on pharmaceutical packaging lines. As a result, machine vision suppliers worked to equip their products to perform optical character verification (OCV) so that their systems could provide the level of performance needed to meet pharmaceutical packagers' requirements. At Itran Corp. (Manchester, NH), a pioneer in the field, we worked to provide the systems that had that performance. I remember, when our engineering department was challenged to provide these solutions, processor designer Mike Greenberg's statement to sales: "You can have better, faster, or lower cost. Pick any two, but you can't have all three."
In the late 1980s, Greenberg's statement turned out to be truer than not. Machine vision systems could add nearly $100,000 to the cost of a new packaging line. While effective algorithms were developed and deployed to perform OCV, they heavily loaded processors, limiting the number of characters that could be inspected or cameras that could be deployed. Operator interfaces were often cryptic and hard to use, and the systems did not communicate well with factory information systems.
John Merva, board member, Automated Imaging Association, (AIA; Ann Arbor, MI)
In 1999, much has changed. It is an extremely dynamic time for the pharmaceutical and medical device industries. Mergers of major players are announced regularly. New drugs are approved at rates never before attained. Medical devices have reached new levels of sophistication. With all of these developments, the number of applications for machine vision systems, and the performance required, is increasing.
Machine vision system performance in packaging applications has improved dramatically. Today's processors offer rapid computational power at a fraction of the cost of their predecessors. A single processor board can now support multiple cameras performing OCV at packaging line rates. Furthermore, that same processor cannot only verify human readable codes but can also read 1-D bar codes and 2-D data matrix codes and perform label skew and cap presence inspections. Advances in lighting and imaging allow tablets and products to be inspected through plastic packages with no increase in rejects from glare or dark spots in the image. Finally, the PC platform has become the host of choice for these systems. Contemporary Windows-based operator interfaces help line operators set up easily and identify problems by summarizing inspection results from all points along a packaging line in one, easy-to-interpret place. Communication to factory information networks is now easily accomplished.
It may have taken 10 years, but guys like Mike Greenberg have given us what we wanted all along: better, faster, and lower cost. The challenge now stands for packaging line manufacturers and users to utilize these systems in an attempt to eliminate product defects and minimize false rejects while increasing machine output.
John Merva is also general manager of RVSI/Northeast Robotics, a New Hampshire-based company that provides illumination and imaging products for the machine vision and 2-D symbology industries. The Automated Imaging Association (AIA) is hosting The Vision Show October 5–7, 1999, in San Jose. For more information, contact AIA at 734/994-6088.
