CD0810_LearnCurve

Seeing the Learning Curve in Color

Sept. 16, 2008
Simplicity Is the Aim in First Color-Sensor Installation
By Phil Burgert

Keep it simple. That’s always a good suggestion for a first-time technology installation. This gets special emphasis when you ask for advice on a first-time, color-sensor installation.

“I prefer a simple design and try to stay to a self-contained housing,” says Rick Bondy, product manager for advanced sensors at Sick. Many sensors now come with teach functions that allow for easy setup that can be performed just by showing the sensor a color to remember or a range of color tolerances, adds Bondy. “So, if want it to see a particular shade of blue, then through your teach function you can say, ‘I want to see only that shade of blue,’ or if you want to see some variations of that blue, you can adjust the tolerances up and down,” says Bondy.

At the same time, machine builders installing color sensors shouldn’t underestimate the complexity of color processing, and they should follow the necessary learning curve before trying to tackle a first application, says Endre Toth, director of business development for Vision Components. “Certain image processing functions are very simple, visual and natural to do in one color model but would be extremely complicated to perform in another color model.”

Steve Nylund, CEO of Delta Computer Systems, notes that a color sensor distinguishes between known colors to recognize the color of an object in a factory setting. “Color sensors are self-contained—they generate their own lighting, making them easy to design in,” he says.

Nylund puts color in three categories that include color-mark readers intended to pick a registration mark on a moving target and typically only distinguish between two widely different colors at high speeds, low-end color sensors that distinguish between three or four different colors—as long as they are not too dark or too close in hue—and high-end color sensors that can be taught to distinguish between 10 or more colors.

“A color camera is a good choice in applications such as checking wire colors,” says Nylund. “But for looking at the average color of an object, cameras are a poor choice. They are not optimized to detect subtle color differences and are inherently more susceptible to noise due to the very small pixel size.”

Color vision sensors almost all have industrial Ethernet connectivity and can send data directly to standard industrial networks such as EtherNet/IP or Modbus TCP/IP, notes Jeff Schmitz, corporate business manager for vision systems at Banner Engineering.

But, unlike consumer electronics, grayscale (black and white) sensing is the norm in industrial automation, Schmitz says. “Color sensing adds complexity and cost and should only be deployed for inspections where color wavelength of light requires analysis. If two different color parts need to be sorted, this is often better done with a grayscale sensor and colored lights or filters.”

Color is a fairly small fraction of the overall machine vision market, notes Ben Dawson, director of strategic development for Dalsa. Part of that is because a lot of what machine vision in the past did involved dimensioning, counting parts or sizing, he notes. “And that’s done happily and less expensively in monochrome,” he says. “When doing gauging and measurement, if you can get away with it, monochrome is more accurate.”

Charlie Strobel, product manager at Panasonic Electric Works of America differentiates the lighting requirements between vision systems and sensors, stating that, “the further away you get, the more light you need on an object, especially for a vision system,” he says. ”Our sensors use their own light source basically to reflect on an object to make sure that the color is correct coming off it. What would be bad for that is if you had intense lighting or something else that would distort that light.”