Originally Published PMPN October 2001
SPOTLIGHT
Evaluating OCV TechnologyEvan Lubofsky, senior writer, Cognex Corp. (Natick, MA)
As demands for greater security and traceability in pharmaceutical manufacturing and packaging have increased, the need for automatic verification of date and lot information has become more important than ever. Human-readable character strings, whether they appear on cartons, bottles, or blister packs, need to be verified with the highest degree of accuracy and reliability, and often at exceptionally high line speeds.
Pharmaceutical manufacturers generally rely on machine vision technology for automatic date and lot code verification. In a typical application, a video camera mounted to a labeler or other piece of packaging equipment captures an image of the character string to be verified, which may or may not be printed on a label. The image is then processed using specialized optical character verification (OCV) algorithms, which compare the read characters to a set of characters previously taught to the system. If an incorrect character is found, the vision system may send a fail signal to a programmable logic controller (PLC), which in turn may trigger some type of reject mechanism to eject the package from the line.
While the role of vision systems in pharmaceutical OCV have remained somewhat consistent over the years, the capabilities of OCV technology, and the types of vision products that are available, have changed. Some manufacturers prefer traditional PC-based solutions, and others are moving toward low-cost vision sensors. And some vision products are designed for those with little or no previous machine vision experience, while others are not. Following are a number of things to consider when evaluating OCV technology.
ROBUST ALGORITHMS
To provide the accuracy and repeatability required for today's demanding pharmaceutical applications, OCV technology should feature algorithms robust enough to verify codes under a wide range of conditions. Labeling equipment, package and container materials, printing methods, and ambient lighting conditions in the plant can vary considerably from one manufacturing facility to another. Today's OCV technology needs to be adaptable to each.
In particular, look for technology that can verify codes repeatably, even when faced with strings containing unevenly spaced characters, confusing characters (such as "B" and "8"), or touching characters. It should also be able to tolerate wide changes in string or label orientation and handle both vertical and horizontal strings. Finally, OCV technology should be designed to analyze each individual character of a string, rather than only performing verification of the string as a whole.
During system evaluation, look at performance under conditions that are more extreme than what the vision system may actually have to contend with. For example, present the system with labels printed at lower contrast than usual, rotate the label a full 360°, use strings with multiple confusing characters, and alter the lighting from high to low. Extending your application requirements during the evaluation phase not only helps assess the robustness of the OCV technology, but may also help reduce the chance of false rejects during production.
STATISTICAL FONT TRAINING
Because normal print can vary a great deal from label to label, printed characters may appear different from string to string. Thus, it is important to be able to train a vision system to handle multiple instances of a string, each with different visual characteristics. Products that offer statistical font training allow you to create a single model from a series of images. This enables the system to better handle the normal range of print quality variations, such as poor contrast, placement, degradations, or variations in stroke widths.
Font training should also be relatively simple and straightforward. To simplify the task, some products offer graphical font training utilities that enable you to train a font by simply drawing a box around the characters to be verified and clicking a train button. Font training utilities may also allow you to see how well characters in a string are segmented from one another as well as to edit trained models to clean up any voids in the print or extraneous marks.
ETHERNET NETWORKING CAPABILITIES
Pharmaceutical manufacturers need a centralized way to maintain and manage the ever-growing number of vision products running throughout a manufacturing facility. Vision products with built-in Ethernet networking capabilities offer a number of advantages. First, Ethernet communications enable you to easily link multiple vision sensors or systems. For example, networking allows inspection setup files to be transmitted over Ethernet to an information database, where they may be stored and later downloaded to other vision systems deployed throughout the facility. Networking may also allow the remote setup of OCV applications as well as centralized monitoring of vision activity. Finally, Ethernet communications can streamline connectivity to PLCs and other automation equipment. Some sensors feature support for EtherNet/IP, a new industrial networking standard for factory device communications that offers direct, seamless links to other equipment, which helps minimize the cost and complexity of implementing OCV on the line.
SECURITY
OCV technology should have a variety of security features built-in. For example, products with password-protected, multilevel-user access may allow only administrators to set up and modify inspections, while restricting operator access to machine operation. A node lock feature, found on some of the latest vision sensor products, enables users to tie a specific OCV application file to a specific sensor by entering in the sensor's unique MAC Address code during setup. This feature ensures that an application configured for one vision sensor is used on that sensor only and not inadvertently loaded onto another.
PART 11 COMPLIANCE
When evaluating OCV technology, make sure the system is compliant with FDA's 21 CFR Part 11 regulations, which relate to technical and procedural compliance for electronic records and signatures and are being aggressively enforced by FDA. Most, if not all, pharmaceutical companies today are moving toward a paperless reporting structure. As part of its compliance initiative, the vision company you're working with should have published application guidelines that provide recommendations for developing Part 11–compliant automation projects that incorporate machine vision technology.
CONCLUSION
In addition to looking for certain specific product features when shopping around for OCV, you may want to work with a vision company that offers a range of OCV solutions. For example, in situations where line speed exceeds 1000 labels per minute and print quality varies to the point where characters' strings appear radically different from label to label, a PC-based solution may work best. However, if the application calls for a lower-cost product that is easier to integrate and set up on the line, a vision sensor may be the solution. Whatever the case may be, having a range of OCV solutions to choose from will enable you to find a product that best meets the specific performance, cost, and functionality requirements of your application.
Perhaps the most important consideration of all is whether a particular product will provide reliable, repeatable, 100% date and lot code verification with no false acceptance. Between the threat of product liability and the watchful eye of FDA on quality control initiatives, there is simply no room for error.
Copyright ©2001 Pharmaceutical & Medical Packaging News
