Tech Talk - November 1999

When the answer isn't in black & white   (continued)
(click on thumbnails to see enlarged images)

Contrast sensors are the most basic type of color sensor and are commonly used for registration mark sensing in the packaging industry. Otherwise known as color mark sensors, these specialized diffuse mode photoelectrics detect the difference in gray-scale response between the registration mark and the background. Color mark sensors are specified by the color of the light source (red, blue or green) which is selected based upon the registration mark and background colors to ensure the highest level of contrast. Selection of the proper light source is simplified through the use of a selection chart, as shown in Figure 6.

Color recognition sensors, however, have the ability to actually learn and distinguish between colors. So, a color mark sensor asks the question "Is there a difference between two colors?" while a color recognition sensor asks, "What color is it?" Also based on the RGB principle, a color recognition sensor contains red, green and blue LEDs (or a single 3-color LED), and sends pulses of each color sequentially (See Figure 7). After a packet of pulses, the sensor analyzes the reflected light for that color and sends pulses of the next color. The information for each set of pulses is then compared to the learned color. This feedback can be based on one sample (single mode) or, when higher precision is required, a series of samples (average mode). In such a case, response time is affected; single sample mode may yield a 1ms response time as opposed to the 10ms required to average 8 samples. Color matching is evaluated by hue and chroma (Color Only mode), or by hue, chroma and value (Color + Intensity mode). The latter mode is ideal for applications involving shiny targets or when a greater level of precision is required, but it is more sensitive to variations in distance to the target object. Color + Intensity mode also allows the sensor to accurately measure color on textured surfaces like lumber, as shown in Figure 8.

To place colors in memory, color recognition sensors are programmed either remotely (through the PLC or controller) or locally, right at the sensor. Programming is simplified through "point and learn," a function that allows target colors to be learned by simply aiming the sensor at the target and pressing a button. Other features might include timing functions like "pulse stretching," which allows slower PLCs to be used in high-speed applications, and gating inputs that enable sensors to be used in tandem for precise target alignment.

With beam patterns and behavior similar to fixed focus photoelectrics, most color recognition sensors employ fiber optic cables. Placing a lens assembly in the fiber optic cable rather than in the sensor housing allows for more flexibility, especially in space-critical applications; the optics can be placed at the sensing location while the electronics can be mounted elsewhere. Signal strength is optimized through the use of bifurcated fiber optics, guaranteeing successful reads regardless of color variance or lens contamination. Standard fiber optic cables may also be used, but with some loss in sensing range and ambient light rejection.

An example -

The most effective way of describing the difference between color mark and true color recognition is to compare how each would see the same target. Figure 9 compares the views of a target as perceived by two color mark sensors (red and green light sources) and a true color recognition sensor. The color marks are seen as they truly exist - red marks on a yellow background - only by the recognition sensor. Since color mark sensors evaluate the reflection of a single color (the light source color), they result is simply a grayscale view of the sample. The best color mark sensor for the red and yellow sample would have a green light source to provide the greatest contrast between the mark and the background.

In summary -

When choosing a color sensor, the key is to understand what information is required for your application. If you simply need to know that a color/registration mark is present, then a contrast sensor is the answer. But, if you need to know the color of that target and the process requires feedback when there are deviations in that color, you need true color recognition.

Thanks to Todd Bissell, Peter Cicchetto, Jeff Holman and Witold Szypszak for their collaborative efforts on this article.