Making the right color match

E-mail

Most machine-vision installations rely on black-and-white images, but increasingly color is becoming a necessity. One color application that is gaining ground is “color matching,” according to Glenn Archer of EPICVision Solutions, a St. Louis integrator that designs machine-vision systems for blue-chip customers like Boeing, 3M, and Procter & Gamble.

“A good example of color matching,” explained Archer, “is building components. Your vision systems help ensure that the color of a new replacement part will match the color in an assembly produced earlier.”

What’s involved in color matching? Archer, an electrical engineer and director of business development, described a system that EPIC designed to inspect 5-ft-long extruded plastic building components on the production line. The setup required lighting techniques that would allow highly consistent and repeatable color detection. Furthermore, the system had to deliver the required accuracy in an environment where temperatures can reach more than 100°F.

To meet these challenges, EPIC built a 3-ft-long enclosed inspection station designed to block out ambient light. Openings just 4-in. high on two sides of the ventilated station permit entry and exit of components as they move along the conveyor. Lighting inside the enclosure consists of two, 18-in.-long white LED strips placed on either side of the part at 45° angles for optimum imaging.

“We initially selected fluorescent light strips but found that we could not achieve the uniform intensity we wanted,” explained Archer.

To capture the images, the engineers chose a 2-Mpixel color PixeLINK camera that connects via FireWire to an adjacent PC for image processing. The camera includes built-in temperature compensation, a key feature since small changes in ambient temperature can adversely affect image consistency. A software utility from the camera manufacturer called “flat field correction” also contributes to uniform lighting intensity by eliminating hot spots.

The software used to program the application, LabView, allowed for a large region of interest (ROI) to be directly converted from the camera’s RGB values to CIELab values, which was the color space that the customer chose for evaluating images (see figure).

System software combines the 15 individual images needed to represent a complete part into a unified average for comparison to a reference color. It also lets operators match a part’s color to that of a part produced previously. With the strategically chosen ROI size, operators can also set pass/fail criteria for each image, based on such anomalies as stains, spots, and streaks.

How much does a color-match vision system cost versus a black-and-white solution? Ballpark estimate: about 20% more, according to Archer. The added expense lies not so much in the hardware but in the extra engineering required to solve tougher application challenges.