Tech Talk - November 1999

When the answer isn't in black & white

(Click on thumbnails to see enlarged images)

This fall, Rockwell Automation, Presence Sensing Business enters the color sensing arena with the introduction of the ColorSight photoelectric sensor. Using the industry standard Series 9000 package style, ColorSight incorporates new technology to bring affordable true color sensing capability to the industrial plant floor. For more information on ColorSight, and other Rockwell color detection solutions, refer to page 32 of this edition of Sensors Today. The following article is a primer on the fundamental theory and operation of color sensing.

Color sensing, though it sounds simple, is likely one of the most difficult sensing technologies to apply and comprehend. While standard photoelectric sensing is based on sensing the amount of light energy, color sensing demands a qualitative analysis of light. Recognition of color implies a virtual consciousness in the sensor–the ability to learn colors and distinguish between their variations. Not all applications necessitate true color recognition; the bulk of color sensing jobs involves detection of differences in reflected light between target and background. Knowing what color sensor to apply in a given situation requires a grasp of light and color attributes, how they are quantified and how the sensor processes this information.

Defining and Measuring Color: Hue, Value and Chroma

When visible (white) light passes through a prism, it breaks up into its component colors (red, orange, yellow, green, blue, indigo, and violet), each with a unique wavelength (see Figure 1). An object's color is based on the properties of the light source, the object and the receiver. As shown in Figure 2, the reflected light is perceived as having a specific color by the observer. Because the light and color receptors in the human eye (rods and cones, respectively) vary from person to person, no two people see colors the same way. Nor would they express color in the same terms. For example, a red sportscar speeds past a group of people. If you were to ask those people the color of that automobile, responses might range from fire-engine red to hot red to just plain red. Several other factors affect color perception: light source and angle, ambient light, background, viewing direction, object size and surface texture. To effectively measure color, however, we use a set of color characteristics and define them numerically. These basic characteristics, or attributes, are hue, value and chroma.

Hue refers to the basic color families as they appear in the visible light spectrum: red, yellow, green, blue and purple. Adjacent colors are then mixed to create a secondary set of hues and, ultimately, a continuous color wheel as illustrated in Figure 3. The wheel is incremented such that a hue can be identified numerically.

The lightness, or value, describes how pale or dark a hue is on a scale of 0 (pure black) to 10 (pure white). Black, white and the gray scale in between are without hue– otherwise known as neutral colors. A good example of how this attribute affects a hue would be lime green versus emerald green; lime is a light (high) value of green while emerald, leaning more toward black, is a dark (low) value of green.

Chroma, also known as saturation, is the deviation of a hue from a neutral color of the same value. Vivid or bright colors have a high chroma while dull or weak colors have a low chroma. Changes in chroma involve adding or subtracting color without changing the value and hue. Neutral gray is considered the chroma origin (zero chroma) with higher numbers expressing increased brightness.

By arranging these attributes as coordinate axes, a three-dimensional 'color space' is developed which allows for numeric expression of color (See Figure 4). While several organizations have developed their own color spaces, industrial color sensors are generally designed and calibrated around a single color space. From that color space, the relative degree of change in a color in terms of hue, value and chroma, can be determined. This change, E, is defined as: , where H equals change in hue,V is a change in value and C is a change in chroma. However, the relative size of E is not the same for all color spaces. For example, a E of 10 in one color space may correspond to a E of 100 in another, but the absolute color change may be the same. The degree of change that a sensor will indicate is also denoted by E, which translates into the resolution of a color sensor. Therefore, when E is specified for a sensor, it is important to note the color space for which it is calibrated.

Color Sensors

Complex color sensors like colorimeters and handheld spectrophotometers provide numeric feedback for even the most minute color differences. However, a significant portion of color sensing applications require only an indication of the presence or gray-scale change of a color mark. Other applications involve true color recognition; a function which requires that the sensor learn a specific color (or colors) and its attributes. A good example of a color recognition application is the sorting of milk jugs according to cap color (See Figure 5).

Industrial color sensors generally operate on RGB (Red, Green, Blue) technology. That is, red, green and blue light are emitted from the sensor and a response is generated based on the amount of each color reflected off the target and received by the sensor. Light sources can be incandescent or LEDs. Incandescent light sources require a series of filters (one each for red, blue and green) and generally have a life span of less than 20,000 hours. As a rule, LED technology is more expensive than incandescent, but has life spans on the order of 100,000 hours. LEDs also allow for reduced sensor size and more effective heat dissipation.

...more