Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org

Volume 18, Number 3 - May 2002 to July 2002

Classifying Colored Bar Codes to Predict Scanning Success By Dr. Mathias J. Sutton, CSIT

KEYWORD SEARCH

Graphic Communications Information Technology Printing Production Quality Control

Reviewed Article

The Official Electronic Publication of the National Association of Industrial Technology • www.nait.org © 2002

1 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org Classifying Colored Bar Codes to Predict Scanning Success By Dr. Mathias J. Sutton, CSIT

Overview accurate information takes too long to The theory and methods of bar process, an organization is forced to Dr. Mat Sutton, CSIT, is an Assistant Professor of In- code print quality verification have operate in a historical mode, missing dustrial Technology at Purdue University, where he teaches courses in Automatic Identification and Data progressed over the last decade. valuable business opportunities. On Capture, Engineering Economy, and Measurement However, there is precious little in the the other hand, information processed and Evaluation in Industrial Technology. He is also a transportation logistics officer in the Air Force Re- literature with respect to colored bar quickly, but inaccurately, causes poor serve. Major Sutton is currently on a military leave code symbols. A search of the Institute business decisions. of absence from Purdue – on active duty in support of Operation Enduring Freedom, where he was re- for Scientific Information Citation Automatic identification and data cently awarded the Meritorious Service Medal. Database revealed only one published capture (AIDC) evolved over the years to article (Agroskin & Golubovskii, 1996) help improve both timeliness and related to the effect of on bar accuracy of information needed to make code symbol contrast. A search of the strategic and tactical business decisions. Automatic Identification Manufactur- Just as logistics provides time and place ers’ (AIM) and Institute of Electrical utility for goods within the supply chain and Electronic Engineers’ (IEEE) (Coyle, Bardi, & Langley, 1996), AIDC jointly sponsored Workshop on Auto- provides time and place utility for critical matic Identification Advanced Tech- data within a company and, in some nologies conference proceedings cases, between supply chain partners. revealed one related paper (Sutton, According to Dunlap (1995), the 1999). In most cases, the literature goal of AIDC is to immediately identify gives general guidelines, but limited physical objects with 100% accuracy research has been conducted to help the and to pass this information to the host practitioner make educated choices on computer for instantaneous decision bar-space color combinations. making. Perhaps the most familiar, and Agroskin and Golubovskii concur, certainly the most predominant, form of stating that information regarding the AIDC technology is bar code. spectral properties of bar codes and the Over the past 30 years, linear bar spatial distribution of reflected is codes quietly infiltrated and permeated “absent in the available literature” (p. the retail, manufacturing, and logistics 229). The purpose of this research is to industries. Their popularity, in part, is a propose a standard method of objec- direct result of the speed and accuracy by tively classifying colored bar codes to which data are entered into a computer help predict, before printing, how well for processing and decision making. In a bar code symbol might perform after simple terms, information is encoded in it is printed. To do this, an older, well- the widths of the bars and spaces. As a established method of measuring color scanner passes over the bars and spaces, will be applied to the field of colored it “senses” the amount of light reflected bar code print quality verification. back to it. The amount of time necessary to scan a given bar or space determines Introduction its width and is converted into digital The lynch pin of any successful information which is meaningful to a organization, operating on the tenets computer (Palmer, 2001). of lean manufacturing or lean logis- tics, is information. Accurate, timely Bar Code Verification information drives the corporate A successful decode depends decision-making process. Accuracy greatly on how well a scanner can and timeliness are both required. If distinguish the dark bars from the

2 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org lighter spaces. In essence, there must be printers of labels and packaging that The Issue of Color sufficient contrast between the bars and employ bar code symbols. Many When ANSI (American National spaces for a scanner to read the printed experts agree that ideally, the blackest Standards Institute, 1990) published the symbol successfully and reliably possible bars should be printed on the Bar Code Print Quality Guideline, they (Collins & Whipple, 1994; Harmon, whitest possible background (Collins & specified the optical geometry and 1994; Palmer, 2001). As bar codes Whipple, 1994; Erdei, 1993; Harmon, general method for verifying bar code evolved since the early 1970s, so too did 1994). Harmon concedes that, in print quality. Their premise was that bar the methods by which to evaluate a practice, this is not the case. A casual codes are primarily printed on symbol’s print quality. The branch of inspection of the items that line the – the ideal situation. They recog- bar code data capture that involves shelves of a typical grocery store bears nized colored bar codes existed – their measuring and evaluating bar code print this out. Bar code symbols are printed method works equally well for black on quality is called bar code verification. on a variety of substrate materials and white symbols or for colored symbols. In 1983 the American National in many different to coincide However, determining all possible Standards Institute’s (ANSI) X3A1 with a package’s . acceptable color combinations was Technical Subcommittee on Optical The issue of color is significant for beyond the scope of the ANSI guideline. Character Recognition (OCR) began those who design labels or packages Why classify bar code colors? It’s studying the issue of bar code print that use bar codes. Although people one thing to know that and quality. The committee’s goal was to can distinguish between most color reflect laser light poorly and, develop a technically sound method of combinations, they interpret colors therefore, perform well as bars and not evaluating bar code print quality that differently than does a scanner (Fox, well as spaces. Where does one draw emulated the operation of a bar code 1991; Palmer, 2001; Stamper, 1989). the line between the color names they scanner (Allais, 1995; Mullen, 1994). As a result, several guidelines are assign to an object? Where is the After numerous meetings and discus- available similar to Erdei’s (1993) and distinction between bluish- and sions over the next seven years, the Bar Stratix’s (1995), which show various greenish-? Assigning color names Code Print Quality Guideline – ANSI color combinations that work best for to an object becomes quite tricky X3.182 was published as the new bar printing bar codes – bar codes that will because of the human perception code print quality method. have enough contrast for successful involved in identifying a colored object. The new guideline was a consider- scanning. For example, Erdei sug- When a person assigns a color name to able leap forward for the science of bar gested four bar colors when illumi- an object, they do it through the lens of code verification. Where the tradi- nated by 633 nm red light: black (most their personal experience, which tional method evaluated bar codes suitable), blue (with high con- understandably varies from individual to according to how they were seen by the tent), green (with low content), individual. Objectively classifying or human eye, the ANSI guideline was and (dark only with low red grouping colors () gives a better written around how a scanner views content). In addition, he suggested indication of how well that will them. Tedesco (1992) described the four background (space) colors: white reflect light. If we know a bar code’s ANSI method as “emphasizing how (most suitable), yellow (very good), (bars and spaces) and well a symbol will scan rather than (with no components from reflectance properties, we can determine how well a symbol is printed” (p. 34). other colors), and red (with no compo- symbol contrast and better predict how Logically, the ANSI method focuses on nents from other colors). well the bar code will scan. several reflectance parameters. Based on general , To bridge the gap between subjec- The first attribute to affect a Erdei’s (1993) and Stratix’s (1995) tivity and objectivity, the Optical symbol’s overall grade is symbol suggestions are helpful and practical. Society of America’s Committee on contrast (Data Capture Institute, 1994; Erdei suggested that one should view (1963) defined the psycho- Palmer, 2001). Symbol contrast the color combinations’ contrast physical aspects of color, which involve ()= − SC RMax RMin measures the “through a Wratten 26 red filter in the relationships between physical stimuli difference in reflectance between the same way a scanner will look at them” and the sensory or perceptual responses dark bars and the lighter spaces. If (p. 131). Harmon (1994) made a to these stimuli. They clarified that these enough contrast exists, the scanner can similar suggestion. He noted that when psychophysical relationships are distinguish between the bars and viewing with a red light source (i.e., definitions, in physical terms, of spaces. Two factors that also affect a helium-neon and visible red laser concepts derived from the subjective bar code symbol contrast are the color diodes), bars printed in red, yellow, human responses to physical stimuli. of bars and spaces and the wavelength orange, reddish-, and reddish- For example, a color can be of light used to scan the symbol. brown will not appear sufficiently defined by matching an object with a different from the white spaces. known spectral distribution to an object Colored Bar Codes with unknown color. The object with a The degree of bar code contrast is known spectral distribution can be a primary concern for designers and expressed in terms of wavelengths,

3 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org which are physical quantities. How- color appearance under daylight A Method of Classifying ever, “matching” the colors of both conditions is how we tend to assign a Colored Bar Codes objects to define equivalence is subjec- color name to objects. We see an apple The colorimetry literature clearly tive – clearly psychological. The as red because it reflects light from the reveals that the CIE system is the method relationship between the two, which red end of the . How- of choice for measuring and classifying defines the color, is psychophysical. ever, if the apple is illuminated with a colors, whether in social science or This psychophysical nature of light source that doesn’t contain “red” industrial-technical disciplines (Commit- color is the basis for the International wavelengths, like sodium light, then the tee on Colorimetry, 1963; Hunt, 1987; Commission on Illumination’s (CIE) apple will appear gray or black. Judd & Wyszecki, 1975; Wyszecki & method of color specification. The CIE The color of light illuminating an Stiles, 1982). The CIE system is a color explains that the stimulus for color is object is fundamental to bar code measurement method independent of any provided by the proper combination of verification. The issue is not how one person. The CIE method produces a a source of light, an object, and an people view the bar code, but how the numerical description that gives an observer (Billmeyer & Saltzman, 1981) bar code scanner “views” it. If a bar unambiguous definition of a color, and attempts to tell us how a is illuminated with different light representing the sensation that color might be reproduced rather than how it sources, then the reflectance from the would have on an average observer. The might be described (Rigg, 1987). bar code will understandably vary method involves a combination of three The basis for the CIE system is the based on the light source. Bar code components: observer, light source trichromatic generalization. The trichro- scanners detect the amount of light (illuminant), and object. matic generalization states that over a reflected from a symbol independently wide range of observation conditions, from human color perception. At the Standard Observer many color stimuli can be matched same time when we view a bar code To define a standard observer, the completely by additively mixing some symbol under some form of white light, CIE used a number of people, who were combination of three primary the scanner illuminates the symbol with not color deficient, to match colors (Wyszecki & Stiles, 1982). The set of a different colored light source, additively under standard viewing primary stimuli can be any three (i.e., presumably a visible red laser scanner. conditions. Each observer viewed a red, green, and blue), as long as they Since bar code colors are designed color stimulus shown on one screen meet one condition: mixing two primary and specified by people, there exists a through a 10° field of view. While stimuli will not produce the third. need to describe and classify colors viewing the stimulus, the observer The CIE system allows us to objectively, according to the way adjusted the intensities of three primary describe color in quantifiable terms. An people see them. Once classified, bar lights (R, G, and B), projected onto an object’s color may be described as a code symbol contrast can be evaluated adjacent screen, until the colors matched specific point, defined by chromaticity to determine the overall effectiveness (Figure 1). The average of these coordinates, within a two-dimensional of a specific color group. results, xλ , yλ , and zλ , determined the . Quantifying object color is One way to define or measure spectral tristimulus values of the important to bar code verification. color is based on spectrophotometry. standard colorimetric observer. Chromaticity coordinates facilitate Spectrophotometry involves a wave- classifying colors into relatively homo- length-by-wavelength measurement of geneous groups according to dominate the light transmitted or reflected by a The CIE adopted several standard wavelength and are necessarily involved sample or emitted by a light source illuminants over the years that corre- in describing or measuring color. (Wyszecki & Stiles, 1982). The output spond to typical situations.

is in the form of a curve, called a Most notable are the D illuminants (D55, Measuring Color spectral distribution. This distribution D65, and D75). These illuminants are Contrary to most people’s thinking, shows how the spectral composition intended to represent daylight at all color is not a characteristic or inherent varies across the visible spectrum (at wavelengths between 300 and 830 nm. quality of an object (Chamberlin & discrete wavelength intervals) when For general use and in the interest of Chamberlin, 1980; Committee on illuminated by a specific illuminant. standardization, CIE recommends using

Colorimetry, 1963; Sharkey, 1991). A spectrophotometer’s measure- D65, which represents average daylight Although people commonly associate ments alone do not tell what a color (Berger-Schunn, 1994; Clulow, 1972; an apple with the color red, the apple is looks like. However, these measure- McDonald, 1997). Illuminant D55 not inherently red. Rather, the apple’s ments provide the basic data – spectral represents yellower daylight ( physical make-up is such that it absorbs reflectance values at each wavelength plus skylight); D75 represents bluer some portions of the visible spectrum over the visible spectrum – from which daylight (north skylight). and reflects others. Because people one can determine the color’s appear- One shortcoming of the set of D normally see objects in daylight, the ance from agreed upon conventions. illuminants is that no method was

4 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org specified for reproducing these lights in tristimulus values of an object, whose λ=780 a laboratory. However, Berger-Schunn color we wish to measure, are calculated Yk= ∑ Sλλλρ y (1994) stated that because the spectral using three variables: a CIE standard λ=380 power distribution of an illuminant is illuminant, a CIE standard observer, and generally only used for calculation, it the reflectance values of the object over λ=780 does not have to exist as a light source. the visible spectrum (Judd & Wyszecki, = ρ Such is the case for classifying colored 1975). Both the CIE standard observer Zk∑ Sλλλ z bar codes. and illuminant values (see Tables 1 and λ=380 2, respectively) are expressed over the Object visible spectrum and are published in where: The last component necessary to most texts on colorimetry (Kaufman & Sλ is the spectral distribution of the measure color is the object itself. The Haynes, 1981; Wyszecki & Stiles, 1982) standard illuminant, CIE determined that if the observer and as well as the published CIE standard. illuminant were constant, an object’s Kaufman and Haynes (1981) related ρλ is the spectral reflectance of the color could be represented in a two- that once the spectral reflectance values specimen, dimensional color space, called the of the specimen (spectrophotometer data) chromaticity diagram (see Figure 2) are known, the tristimulus values are xyλλ,,and z λare the spectral widely described in the literature calculated using the following relations: tristimulus values of the standard (Clulow, 1972; Committee on Colorim- observer, etry, 1963; Hunt, 1987; Judd & λ=780 Wyszecki, 1975; Wyszecki & Stiles, Xk= ∑ Sλλλρ x = 100 1982). The spectral locus (the curved λ=380 k is a normalizing factor, ∑ Syλλ boundary) represents a specific wave- length – a color in its richest or purist form. The straight line at the lower right is called the purple line because it Figure 1. Principle of Trichromatic Color Matching by Additive Mixing of Lights. does not represent a unique wave- Red, Green, and Blue are lights whose intensities can be adjusted; Color is the light length. Rather, colors along the purple whose color is to be matched (adapted from Hunt, 1987). line are some combination of red and blue. Near the center of the diagram is Aperture a noticeable white region. At the Plate Red center of the white region is a point whose coordinates represent the Green Diffuser standard illuminant, or equivalently, the Blue Diffuser point where R, G, and B are mixed in Color Observer's equal quantities to form white light. Eye An object’s color can be located within the chromaticity diagram once its spectral distribution is determined and chromaticity coordinates are Figure 2. The CIE Chromaticity Diagram. Colors are represented by coordinates (x, calculated. When using commercially y) within the region bounded by the spectral locus and the purple line (adapted from available colors, like (1996) Chamberlin & Chamberlin, 1980). colors, often the chromaticity coordi- nates are already calculated. If chro- maticity coordinates are unknown, they can be calculated using the data from a Spectral spectrophotometer. Locus Green Calculating Chromaticity Coordinates When viewing the chromaticity Yellow y diagram, all visible colors fall within the Cyan region bounded by the curved spectral Red locus and the purple line (Figure 2). Furthermore, all colors have a unique Purple combination of tristimulus (X,Y,Z) Blue Line values that describe them. The x

5 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org

λ represents wavelengths over the visible spectrum. Table 1. Standard Observer Tristimulus Values (Kaufman & Haynes, 1981).

Given a specimen’s tristimulus Wave-length (nm) x y z values, X, Y, and Z, the chromaticity λ λ λ coordinates for that specimen, x, y, and z may be calculated using the follow- 380 0.0002 0.0000 0.0007 ing equations: 390 0.0024 0.0003 0.0105 400 0.0191 0.0020 0.0860 X 410 0.0847 0.0088 0.3894 x = X + Y + Z 420 0.2045 0.0214 0.9725 430 0.3147 0.0387 1.5535 Y 440 0.3837 0.0621 1.9673 y = X + Y + Z 450 0.3707 0.0895 1.9948 460 0.3023 0.1282 1.7454 470 0.1956 0.1852 1.3176 Z z = 480 0.0805 0.2536 0.7721 X + Y + Z 490 0.0162 0.3391 0.4153 500 0.0038 0.4608 0.2185 The above equations imply x + y + z = 1 and the coordinates are simple 510 0.0375 0.6067 0.1120 proportions of X, Y, and Z that 520 0.1177 0.7618 0.0607 represent the tristimulus values of a 530 0.2365 0.8752 0.0305 particular color. By convention, 540 0.3768 0.9620 0.0137 chromaticity is expressed as x and y, which together correspond to domi- 550 0.5298 0.9918 0.0040 nant wavelength and purity (Judd & 560 0.7052 0.9973 0.0000 Wyszecki, 1975). The third character- 570 0.8787 0.9556 0.0000 istic required to adequately define a 580 1.0142 0.8689 0.0000 color is brightness or luminance. Billmeyer (1981) explained that in 590 1.1185 0.7774 0.0000 the CIE system, Y is known as the 600 1.1240 0.6583 0.0000 luminance factor and represents the 610 1.0305 0.5280 0.0000 perceived of an object. The 620 0.8563 0.3981 0.0000 value Y = 100 is assigned to a perfectly white object that reflects 100% at all 630 0.6475 0.2835 0.0000 wavelengths and is the maximum value 640 0.4316 0.1798 0.0000 Y can have. Chromaticity coordinate z 650 0.2683 0.1076 0.0000 is not used to describe color. Knowing 660 0.1526 0.0603— 0.0000 Yxy is sufficient to describe a color because these values provide enough 670 0.0813 0.0318 0.0000 information to calculate z, and hence, 680 0.0409 0.0159 0.0000 X, Y, and Z – the tristimulus descrip- 690 0.0199 0.0077 0.0000 tion of a unique color. 700 0.0096 0.0037 0.0000 Classifying Bar Code Colors 710 0.0046 0.0018 0.0000 In order to objectively assign a hue 720 0.0022 0.0008 0.0000 (color name) to a bar code, the CIE 730 0.0010 0.0004 0.0000 chromaticity diagram can be divided into 740 0.0005 0.0002 0.0000 conveniently sized regions (see Figure 3). First, determine the size interval and 750 0.0003 0.0001 0.0000 determine the boundaries of those 760 0.0001 0.0000 0.0000 intervals over the visible spectrum (380 770 0.0001 0.0000 0.0000 nm to 780nm) and determine the 780 0.0000 0.0000 0.0000 corresponding chromaticity coordinates

6 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org

(see Kaufman & Haynes, 1981). Next, calculate the chromaticity coordinates (x, Table 2 – CIE Spectral Power Distributions of Standard y) for the bar code (bars and spaces), if (Kaufman & Haynes, 1981). not readily available. Finally, transform the interval boundaries and bar/space Wavelength l(nm) S x S y S z chromaticity coordinates to polar λ λ λ λ λ λ θ coordinates (r, ). Unlike (x, y) coordi- 380 0.001 0.000 0.003 nates which represent a point in a two- dimensional Cartesian plane, polar 390 0.011 0.001 0.049 coordinates represent the same point as 400 0.136 0.014 0.613 the distance from the origin, r, and the 410 0.667 0.069 3.066 angle measure, q, about the origin with 420 1.644 0.172 7.820 respect to the x-axis. The transformation 430 2.348 0.289 11.589 is based on standard illuminant D as the 65 440 3.463 0.560 17.755 origin, instead of (0, 0). Table 1 shows 450 3.733 0.901 20.088 that the chromaticity coordinates for D65 (x , y ) 460 3.065 1.300 17.697 are D65 D65 = (0.3138,0.3309). For example, the following two 470 1.934 1.831 13.025 steps illustrate how the Pantone Green 480 0.803 2.530 7.703 specimen’s chromaticity coordinates, 490 0.151 3.176 3.889 (x , y ) = (0.1872,0.4514), were g g 500 0.036 4.337 2.056 converted to polar coordinates: 510 0.348 5.629 1.040 ( ) ( ′ ′ ) Convert xg , yg to xg , yg such that 520 1.062 6.870 0.548 standard illuminant D65 is the new origin. 530 2.192 8.112 0.282 540 3.385 8.644 0.123 (x′ , y′ )= (x − x , y − y ) 550 4.744 8.881 0.036 g g g D65 g D65 = ()0.1872 − 0.3138,0.4514 − 0.3309 560 6.069 8.583 0.000 = ()− 0.1266,0.1205 570 7.285 7.922 0.000 580 8.361 7.163 0.000 590 8.537 5.934 0.000 ( ′ ′ ) ( θ ) Convert xg , yg to rg , g , the Pantone 600 8.707 5.100 0.000 Green specimen’s polar coordinates. 610 7.946 4.071 0.000 620 6.463 3.004 0.000 = ′2 + ′2 rg x g y g 630 4.641 2.032 0.000 = ()()− 0.1266 2 + 0.1205 2 640 3.109 1.295 0.000 650 1.848 0.741 0.000 = + 0.01603 0.01452 660 1.053 0.416 0.000 = rg 0.1748 670 0.575 0.225 0.000 680 0.275 0.107 0.000 690 0.120 0.046 0.000  ′  700 0.059 0.023 0.000 yg θ = arctan  710 0.029 0.011 0.000 g  x′   g  720 0.012 0.004 0.000  0.1205  = arctan  730 0.006 0.002 0.000  − 0.1266  740 0.003 0.001 0.000 θ ≈ ° g 136 750 0.001 0.001 0.000 760 0.001 0.000 0.000 770 0.000 0.000 0.000 θ Once is determined from D65 to the Sums (X, Y, Z) 94.825 100.000 107.381 spectral locus on the chromaticity Chromaticity diagram for each of the chosen wave- (x, y, z) 0.3138 0.3309 0.3553 length boundaries, the regions are

7 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org defined. A bar code’s colors (bar or would otherwise appear to be the same Allais, D. C. (1995, October). U.P.C. space), whose polar coordinates fall color to individuals viewing them. symbol quality specification. Paper within a specific color region, is Proposing a method to classify presented at the SCAN-TECH, assigned that hue. Given a population colored bar codes to predict scanning Chicago, IL. of colors, like Pantone (1998), plotted in success, suggests further research to American National Standards Institute. the chromaticity diagram, define regions answer the following questions: (1990). American national standard sufficiently small such that variation 1) What bar code colors are generally for information systems — Bar code within the region is minimized. Calcu- preferred for bars? For spaces? print quality — Guideline (ANSI late mean reflectance for the population 2) Is there an ideal bar-space color X3.182-1990). New York: Author. colors that fall within each region. After combination to optimize scan- Berger-Schunn, A. (1994). Practical color we know from what regions a particular ning success? measurement (Max Saltzman, Trans.). bar or space is located, symbol contrast 3) Do the wavelengths of laser light New York: John Wiley & Sons. can be closely estimated by subtracting generally used in bar code scanners Billmeyer, F. W. J., & Saltzman, M. the bar-region mean from the space- significantly affect scanning (1981). Principles of color technol- region mean. This estimate of symbol success of colored bar codes? ogy. (2nd ed.). New York: John contrast will indicate how well the bar- 4) What effect, if any, does the Wiley & Sons. space color combination will perform as saturation level (white compo- Chamberlin, G. J., & Chamberlin, D. a printed bar code symbol. nent) of a hue have on bar code G. (1980). Colour: Its measure- scanning success? ment, computation and application. Discussion London: Heyden & Son Ltd. This research proposes a standard Answers to these questions will help Clulow, F. W. (1972). Colour: Its method of objectively classifying bar provide a bar code with a readable principles and their applications. codes printed on colored substrates so combination of bar-space colors before London: Fountain Press. that scanning success can be predicted. the symbol is actually printed. Further- Collins, D. J., & Whipple, N. N. Current practices follow rule-of-thumb more, addressing these questions will (1994). Using bar code: Why it’s guidelines or the subjective evaluation fill a noticeable void in the bar code taking over. (2nd ed.). Duxbury, of the package designer, if at all. print quality body of knowledge. MA: Data Capture Institute. The primary contribution of the Committee on Colorimetry. (1963). method proposed here removes the References The science of color. Washington, subjectivity of classifying colors, Agroskin, L. S., & Golubovskii, Y. M. DC: Optical Society of America. especially when hues are a mix be- (1996). Microspectrophotometric Coyle, J. J., Bardi, E. J., & Langley, C. tween one that is generally acceptable study of bar codes. Journal of J. (1996). The management of (i.e., yellow) and one that is poor (i.e., Optical Technology, 63(3), 229-231. green). Using the CIE chromaticity diagram to determine a hue’s chroma- ticity in two-dimensional space Figure 3. The CIE 1964 Chromaticity Diagram with Six Color Regions Defined eliminates the variability between two (Wavelength Followed by Chromaticity Coordinates). or more individual’s perceptual classifi- cations of the same color. A standard method is needed to prevent unsuitable (unreadable) bar-space color combina- tions from going to press. Conclusion To date, no standardized method exists to objectively classify a bar code’s bar-space color combination. The proposed method provides a theoretical, yet practical, model that will help determine if a bar code will scan successfully prior to printing. The method serves to substantiate existing rule-of-thumb guidelines. Further- more, the method allows for classifying colors in hue regions that are suffi- ciently small – small enough to differentiate between two hues that

8 Journal of Industrial Technology • Volume 18, Number 3 • May 2002 to July 2002 • www.nait.org

business logistics. (6th ed.). Minne- Kaufman, J. E., & Haynes, H. (Eds.). Sharkey, F. M. (1991, November). apolis, MN: West Publishing. (1981). IES lighting handbook: Basics of bar code scanning. Paper Data Capture Institute. (1994, Septem- Reference volume. New York: presented at the SCAN-TECH 91, ber). Bar code systems: Only as Illuminating Engineering Society of Dallas, TX. good as the poorest bar code. North America. Stamper, B. (1989, October). No Automatic I.D. News, 10, 32-33. McDonald, R. (Ed.). (1997). Colour verification? That’s too high a risk Dunlap, D. D. (1995, October). Physics for Industry (2nd ed.). West to take. Automatic I.D. News, 5, 33. Understanding and using ADC Yorkshire, England: Society of Stratix. (1995). Guide to color selec- technologies. Paper presented at the Dyers and Colourists. tion. Atlanta, GA: Author. SCAN-TECH 95, Chicago, IL. Mullen, D. (1994, May). ANSI alpha- Sutton, M. J. (1999, October). The Erdei, W. (1993). Bar codes: Design, bet soup: A, B, C...X3.182. Auto- effect of incident wavelength on bar printing and quality control. New matic I.D. News, 10, 40. code symbol contrast. Paper York: McGraw-Hill. Palmer, R. C. (2001). The bar code presented at the Workshop on Fox, R. (1991, 4 Nov 91). Bar code book. (4th ed.). Peterborough, NH: Automatic Identification Advanced basics. Paper presented at the Helmers Publishing. Technologies, Summit, NJ. SCAN-TECH 91, Dallas, TX. Pantone. (1996). Pantone color drive Tedesco, J. (1992, August). How bar Harmon, C. K. (1994). Lines of (Version 1.5.3). Carlstadt, NJ: Author. codes can make the grade. Auto- communication. Peterborough, NH: Pantone. (1998). Pantone color formula matic I.D. News, 8, 34. Helmers Publishing. guide. Carlstadt, NJ: Author. Wyszecki, G., & Stiles, W. S. (1982). Hunt, R. W. G. (1987). Measuring Rigg, B. (1987). Colorimetry and the Color Science: Concepts, methods, colour. Chichester, England: Ellis CIE system. In R. McDonald (Ed.), quantitative data and formulae. (2nd Horwood Limited. Colour physics for industry (pp. 63- ed.). New York: John Wiley & Sons. Judd, D. B., & Wyszecki, G. (1975). 96). West Yorkshire, England: Color in business, science, and Society of Dyers and Colourists. industry. (3rd ed.). New York: John Wiley & Sons.

9