Whether for quality control, sorting, or

Transcription

1 Whether for quality control, sorting, or identification, color sensing is a critical part of many automation procedures. Color detection has various meanings depending on the user, including recognizing variously colored plastic pieces to be used in an assembly, confirming the correct sequence of wire colors attached to a connector, or sorting everything from textiles to pharmaceutical capsules. Basic color sensors often have limited functionality and flexibility. This may suffice for some color detection requirements, but when it does not, many users are then forced to consider high-end (and high-priced) color vision systems. Color vision systems may not only mean a cost that is orders of magnitude above that of color sensors, but they also carry the expense of dealing with very customized software, set-up, and auxiliary lighting. However, a more powerful generation of color sensors is emerging to bridge the gap between these two extremes. This advanced series of color sensors has extended detection capabilities and multiple methods to evaluate colors when compared to standard color sensors, but it eliminates the complexity and cost of additional lighting and customized set-up and software. Advanced color sensors, unlike their simpler cousins, learn and evaluate colors much like human vision. They can evaluate colors using an additive approach, looking at each color as a component of high frequency colors, medium frequency colors, and low frequency colors. The standard color sensors often evaluate only one frequency band of light and do not account for the brightness of the color being inspected. See figures 1a and 1b. tristimulus value tristimulus value wavelength nm Figure 1a red green blue wavelength nm Figure 1b red green blue FIgure 1: Color sensors use different colored LEDs or a white LED to evaluate an object. Figure 1a depicts how a standard color sensor can utilize its LEDs to evaluate colors. Note that most object colors are evaluated by only a single color LED of the sensor. Figure 1b depicts how an advanced color sensor makes use of its LED(s)to evaluate an object s color. Note that 2- or 3-color LEDs define an object s color by an additive approach. Before examining advanced color sensors, it makes sense to first cover some basics of colors. Colors can be uniquely identified by coordinates in what is known as a color space. There are different types of color spaces that each defines its coordinates using different means. Most color spaces are three-dimensional. It may make sense to evaluate colors in one color space for one application and to evaluate colors in a completely different color space, with its own uniquely derived coordinates, for another application. A color space not only assigns a numeric identity to a color to uniquely define it, but it can also identify the distance between multiple colors in a color space, which can be helpful to gauge how similar or different one color is compared to another. 1

2 -a -b +L -L +b +a this case the clear plastic blister over the pill) to be considered independently so the color s hue (the pill or capsule) is evaluated separately from its glossiness. Singling out an object s brightness is also helpful because fluctuations in a target s brightness can be ignored if desired. For example, for a standard color sensor, it is often necessary to slightly angle the sensor s lens when looking at very glossy targets, as the glare from the reflected light can blind the sensor and render it unable to detect the color. Using advanced color sensors with selectable color spaces means the sensor can be directed perpendicularly to the target (removing the need for angling) by evaluating the brightness of the target separately. Using a basic color sensor does not offer this level of flexibility, as it operates in one color space, typically XYZ space. Figure 2: A graphic representation of a color space. This 3-D color space is called Lab space. Lower level color sensors often use what is known as XYZ color space to identify colors. XYZ color space was introduced by the International Commission on Illumination (CIE). This color space identifies a color by three values (X,Y,Z) that are derived from the amount of red in the color, the amount of green in the color, and the amount of blue in the color, respectively. More advanced color sensors allow the user to select a color space that is best suited for the application. An example of a potentially useful color space is Lab space, which, like XYZ space, is from the CIE. Lab color space classifies the L coordinate as the lightness/darkness of the color, a as the red-green color component, and b as the blue-yellow component. See figure 2. The first advantage of advanced color sensors is the ability to choose the optimal color space for the application. For example, if it is necessary to differentiate colors of pharmaceutical pills and capsules inside shiny clear blister packaging, it may be helpful to use the CIE Lab color space. Lab space allows the glossiness of the target (in Identifying colors by means of a color space also allows the use of tolerance values. Tolerances allow a space to be mapped out around a taught color to allow for normal variances that can occur when evaluating a color. Variances can occur from slight differences in how successive parts are made, how parts are positioned for inspection, or even imperfections in the objects themselves. Tolerances make it possible to adjust how sensitive the sensor is to a color. By decreasing the tolerance, the color sensor is better able to distinguish very similar colors. By increasing the tolerance, the color sensor is less touchy to variations in the target color and allows more slop or variance to be detected. In terms of color space, tolerances can be visualized as follows. Colors are identified as points in the three-dimensional space. A tolerance is identified as a 3-dimension shape around each point, most commonly a sphere or cylinder. The smaller the sphere or cylinder is, the smaller the tolerance, and the color sensor is more discriminating when deciding if an observed color is the same as the taught color. When the sphere or cylinder is larger, the result is more slop and the sensor ignores color variations. 2

3 Figure 3: A software screenshot for an advanced color sensor. A 2-D slice of a color space is shown with taught color values (depicted as the marks at the center of the ovals) and tolerances for the values (depicted as ovals). For example, when teaching a color sensor what an acceptable color is for a quality control check of a certain plastic molded part, it is helpful to identify a tolerance for that taught color so slight variations can be ignored. However, when evaluating two molded parts, both very similar hues of dark brown, it is helpful to decrease the tolerance around each color so that the two tolerances do not overlap with each other. See figure 3. One obvious advantage of some higher-level color sensors is the number of different colors that can be taught. Basic color sensors can be taught generally one to five different colors. But most advanced color sensors are capable of recognizing twenty or more different colors, sometimes even several hundred colors, by one sensor. This allows the sensor to easily adapt if additional colors are added to the application in the future. Another strength of advanced color sensors is the ability to detect not just passive objects such as plastic parts, colored glass, cloth, packaging, etc., but also the ability to identify objects that emit colored light. This includes objects such as LEDs on car dashboards, pilot lights, and stack lights. Not only can color sensors confirm that the selfilluminated object being evaluated is emitting light, but it can also confirm that the emitted light is the desired color, thus performing two tasks in one. The flexibility of advanced color sensors also means that it can be possible to forgo using tolerances altogether in favor of an evaluation method called classification. Classifying colors means that any color that is evaluated will be associated with the closest color already taught in color space. An example of this is a sorting application in which it is known that ten different color totes will pass on a conveyor belt, and it is desired to detect the color of each passing tote. Perhaps some totes are older and more worn, so their colors will be slightly faded from a newer tote. Classification means that after the ten tote colors are taught to the sensor, every subsequent tote seen is linked to the nearest of the ten taught colors in color space. So a slightly faded brown tote will still be seen the same as a new brown tote. See figure 4. 3

4 Figure 4: A software screenshot for an advanced color sensor. Colors are evaluated in classification mode, in which the output nearest an already taught color value is always activated. Note classification mode does not utilize any color tolerances. Many more sophisticated color sensors also allow any colors to be grouped together to specific output combinations. For example, if it is desired just to separate sections of carpeting into three groups: beige types, gray types and black types, but there are a dozen different types of each, grouping can be a simpler and faster way to automate the sorting process. In this case, the multiple samples of the gray carpet group are taught to the sensor and are all linked to the same group 1. So whenever one of the gray carpet samples is seen, then output #1 turns on. Multiple beige colors are all taught and linked to group 2, and likewise, whenever a beige carpet sample is present, then output #2 changes state. Some advanced color sensors also offer the possibility of verifying that a correct sequence of colors is seen. This can be critical when inspecting color-coded wires going into connectors and assemblies. In this case, the sensor is taught colors in a particular sequence and the sensor is told by the user to operate in sequential mode. Only when the colors are seen in the proper sequence again does the sensor s output change state. Often an external trigger, such as from another sensor, or an internal timer can be used to determine when the next color in the sequence should be seen. This is an easy quality check to verify that an assembly or connector is put together in the correct fashion. See figure 5. Fiber optic cables complement many color sensors, ensuring color identification can be done under more challenging conditions. Tight sensing confines, extreme temperature conditions, and small objects to be evaluated make fiber optics a better overall choice for many applications. Fiber optic cables are available in diffuse (or bifurcated) mode, in which the transmitter and receiver elements are both located on one single fiber optic cable tip, or in thru-beam (or opposed) mode, in which two separate fiber optic cables are used, one as a transmitter and one as a receiver. Most color applications use diffuse mode fiber optic cables, as the object being evaluated would provide 4

5 an adequate reflective surface to send transmitted light back to the receiver elements in the fiber optic tip. But if the material being evaluated is semi-transparent, such as with some liquids, films, and wrappers, then a thru-beam mode fiber optic cable is the best choice. Because most of the light passes through a transparent object, the thru-beam fiber optic cable doesn t require light to be reflected from the object, but actually requires some light to pass through the object. Fiber optic cables are typically available with a range of different tips to fit the mounting requirements of many applications. Color sensors often utilize complex algorithms and routines to compare an object s color against many already-taught colors looking for a match. Using a color sensor with a rapid response time is instrumental to ensuring improved throughput. Also, using software that is plug and play eliminates the unpleasant task of creating and/or customizing software as would be done for a color vision system. Color sensor software should be both user-friendly and intuitive to use. Graphical displays provide an invaluable tool in quickly ascertaining the similarity of colors to each other and if the defined tolerance is suitable. Color sensors have evolved to offer more control and more flexibility without the drawbacks of a fullfledged vision system. The benefits outlined offer a bridge between the shortcomings of simpler color sensing technology and the pitfalls and expense of color vision systems. Figure 5: Multi-colored wires in a connector. Advanced color sensors can be used to verify that the wire color code follows the correct sequence. Jeff Allison Product Manager Pepperl+Fuchs Twinsburg, OH fa-info@us.pepperl-fuchs.com 5