Lighting Takes Center Stage in Pharmaceutical Inspection

Any engineer will tell you that a successful machine vision application is 90 percent lighting. And anyone working on pharmaceutical inspection, which presents difficult-to-image surfaces like shiny blister packs and glossy labels, couldn’t agree more.

‘‘By far the most critical component [of pharmaceutical inspection] is illumination,’‘ says Jeff Snyder, packaging business development manager for RVSI Acuity CiMatrix (Nashua, New Hampshire, USA). ‘‘All machine vision applications on a reflective surface demand the same thing: proper illumination.’‘

Machine vision helps pharmaceutical inspection on three fronts: analyzing quality, distinguishing flaws, and ensuring the accuracy (and traceability) of data. Poor illumination could feasibly affect all three aspects. Operations employing flexible, highly reflective materials have pushed the limits of machine vision technology. That’s why knowing how to appropriately light the variety of parts in pharmaceutical inspection is crucial to its success.

‘‘The geometry of the object being inspected must be analyzed according to five properties: absorption, texture, elevation, shape, and translucency,’‘ says Snyder. ‘‘When we consider a highly reflective surface, absorption (which dictates how the surface reflects light) and texture are very important. For example, a flat specular surface would benefit most from diffuse coaxial illumination, whereas a product like RVSI/NER’s Square Continuous Diffuse Illuminator (SCDI) is most appropriate for shiny, slightly rounded surfaces like blister packs.’‘

The shape, or shapes, of an object will also affect the lighting method. ‘‘Sometimes with pharmaceutical parts you get into more complex orientations than just a single sheet of foil,’‘ says Greg Hollows, imaging product line manager for Edmund Optics (Barrington, New Jersey, USA). ‘‘There might be layers of different topographies.’‘ Consider bags and pouches, for example. A label marked on the bag carries quality control, tracking, and other information. While flat when empty, these flexible packages expand and change shape as they are filled, sealed, and moved. ‘‘Not even the best illumination system can produce consistent images when subjected to random changes in an object’s appearance,’‘ says Steve Cruickshank, principal product marketing manager, PC Vision, with Cognex Corporation (Natick, Massachusetts, USA).

So how can a vision system acquire an image of that information—printed on a wrinkled, warped, or cured surface—and convert it to useful data? How can the vision system check the printing on the bag to ensure it meets package quality and safety standards? ‘‘When lighting can’t deliver consistent image formation,’‘ Cruickshank explains, ‘‘it really comes down to the robustness of the software.’‘ 

Cognex, for example, offers a software tool called PatFlex to address problems such as this caused by the increased use of non-rigid packaging. PatFlex is designed to recognize distorted patterns to locate an object, feature or pattern whose perspective or appearance changes randomly. PatFlex corrects the image’s severe perspective and spatial distortions so that traditional pharmaceutical applications such as optical character verification (OCV), barcode, and print quality inspection tools can then be applied.

Considering the System’s Optics
The U.S. Food and Drug Administration’s CFR 211 regulation governing pharmaceutical inspection requires 100% inspection and traceability of product and is another major concern for pharmaceutical manufacturers, one that cannot be addressed easily without machine vision. Pharmaceutical manufacturers label their products with character strings that represent factory, date and lot codes, and other information to ensure greater security and product traceability. But the date and lot coding itself can be difficult to image if the appropriate lighting solution is not used.

‘‘Many pharmaceutical applications involve OCV on a label or carton. Many times, the label and carton are coated with a reflective varnish,’‘ says Snyder. ‘‘OCV inspections are also performed on the highly reflective foil sealing material used for blister-packed products. Inspection of tube products can include OCV as well. The medical device industry is beginning to utilize direct-part marks…to provide a permanent identification code on products.’‘

‘‘In many cases, highly reflective surfaces can be challenging to mark,’‘ Snyder continues. ‘‘This means a printed code may smear slightly, or possibly have reduced contrast. When performing an OCV inspection on this type of surface, OCV algorithms must allow process variations but still be able to detect unreadable characters.’‘

Varying topographies and surface textures come into play here as well, adds Cruickshank, such as when the characters or coding run across a seam. ‘‘The biggest challenge is not only that the object is shiny but not necessarily consistent within the image,’‘ he says. ‘‘There’s a gray background and lighter gray text…part of it is on a smooth shiny surface, and part of it is on a textured surface, even within one character.’‘

While proper illumination can minimize specular reflection and hot spots in an image, character strings become distorted and difficult to verify for other reasons, such as the printing process. When printing on moving objects, skew distortion is common due to misalignment of the print head with the direction of product motion. Similarly, strings can become compressed or stretched when the line rate of the marking system is not properly synchronized with the speed of the product as it moves past the printer. Strings marked on curved surfaces tend to have perspective or bow distortions.

Optics play an important role, too. After lighting, Philip Colet, vice president of sales and marketing for Coreco Imaging (Saint-Laurent, Quebec, Canada), suggests, ‘‘The second part of the equation is the optics you’re using and the kind of filters you might be running with as well. If you’re dealing with IR or single-wavelength light to get a better image, it’s really going to start with the lighting and optics you’re using in order to create a separation from the background and the foreground.’‘

Considering the optics of the system can in turn help with the lighting solutions. ‘‘In a lot of cases, it’s about being able to employ other optical components into the mix to get lighting needs taken care of,’‘ Edmund Optics' Hollows says. Examples include adding a diffusing material in front of the light; using multiple light sources or different lighting techniques to accommodate changing conditions within or surrounding the application; using structure illumination like lasers; using mirrors to help redirect light to where you want to have it; and masking part of the source with blinders or enclosures.

The purpose of all this is to increase contrast between the region of interest (ROI) and the background, whether that be the part itself or material handling equipment. ‘‘There’s a very slow gradient in image as you move from left to right, going from darker to lighter,’‘ says Colet. ‘‘A single threshold applied on that image would just give you useless data; a locally adaptive threshold would actually change a value as you move the image. Because of this, you’re still able to distinguish the foreground from the background, regardless of shading or contrast in the image.’‘

In the end, the robustness of the machine vision system will be just as important as the lighting. ‘‘When you’re looking at a lot of these foil packs, you’re looking at highly reflective cellophane-type material,’‘ says Mark Sippel, In-Sight product marketing manager for Cognex. ‘‘It’s always going to start with a good lighting technique, but the robustness of tools will come into play. These packages are so variable in their appearance, that the vision tools need to be able to handle nonlinear lighting and other types of variation.’‘

Despite the obstacles in imaging highly reflective parts in the pharmaceutical industry, machine vision offers a solution. ‘‘When I first started in the vision industry [17 years ago], I had to become proficient at creating my own custom lighting solutions,’‘ Snyder says. ‘‘Today, companies have standard products that solve most challenging applications. If a standard product does not exist, a custom solution can be designed and built.’‘