InnovativeApplications
Understanding color
machine vision
From simple sorting to complex analysis of print, color vision systems can provide
production verification if the proper controls are built into the system.
By Steven Prehn

Many color-verification systems match the color contained within images to a predefined color. Determining the extent of color control needed in the system depends upon the application and how the tools are applied. If a part simply needs to be verified as red and the system need not analyze the exact color composition, strict controls may not be needed. However, if a machine-vision system needs to ascertain that the red exactly matches a defined pantone color, the system may benefit from white-balancing and color calibration. Special attention should also be given to control even minor color shifts that are the result of aging lights, temperature, and ambient lighting conditions.

Blue (0,0,255) Black (0,0,0)

Cyan
(0,255,255)

Magenta
(255,0,255)

Shade of gray

White
(255,255,255)

Gre (0,25

en 5,0)

FIGURE 1. Typical 24-bit color cameras separate images into individual red, blue, and green planes, each with 8 bits of depth using color filters. By mixing different R, G, and B values, all visible colors can be created (top). HSI space is rendered by turning the RGB color cube on its black and white axis. Hue is then represented as the color on the outer rim of the circle and saturation is the white component of the color as represented by the distance from the center of the color (right).

Red
(255,0,0)

Yellow
(255,255,0)

Saturation

Hue

When designing a color vision
system, the wavelengths contained
in the light source, those reflected
and absorbed by the surface of the
part, and the frequency-response
curve of the camera must be consid-
ered. Like the eye, a camera gathers
light reflected (or transmitted) from
objects. If light strikes a shiny surface and fuse or disparate light sources such as ring Like monochrome systems, color ma-
reflects into the camera, surface color infor- lights. Fortunately, several companies pro- chine-vision systems operate effectively if
mation is lost. For this reason, color vision vide products that improve the robustness they can easily detect the differences be-
systems will often benefit from using dif- of color systems. Most often, color machine- t ween good and bad parts. Choosing the
vision systems should use cameras that pro- correct color space is an important aspect

STEVEN PREHN is business manager, vide automatic white balancing, and lights of the system design, as it will enhance
aptúra Machine Vision Solutions (Lansing, that provide a uniform distribution of in- the separation distance applied to the
MI, USA; www.aptura.com). tensity over the color spectrum. colors. Pure white light is comprised of a