Color Effectiveness of Yellow Pavement Marking Materials: Full Report (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

18 CHAPTER 4: DARK-ROOM REAR-PROJECTION SCREEN EXPERIMENT This experiment was conducted in a dark hallway with a 4ft by 7ft calibrated rear-projection screen and projector located at the end. The purpose of the experiment was to determine color classification responses for a common pavement marking viewing geometry. The experiment simulated 4” wide (10cm) continuous yellow pavement marking centerline stripes laid out on a straight and level roadway with 12ft (3.65m) lane width, with the markings gradually fading into white along the centerline stripe. Pavement marking luminance, pavement luminance, viewing geometry (gaze direction), and pavement marking chromaticity were controlled in a full-factorial design. For each condition, each participant evaluated the general color appearance of the pavement marking stripe on the rear-projection screen as either “yellow” or “white”. During the evaluation, the response time between the onset and the response was measured. Also, each participant indicated the location of the transition point from yellow to white on the continuous centerline stripe with a mouse pointer. The chromaticity of each transition point was also recorded. The results are summarized in the Results section starting on page 22. METHOD Yellow pavement markings were presented on a back-projection screen using an EIKI (LC- SX1U) LCD projector. The pavement marking stripes were saturated yellow at close distances and desaturated from yellow towards white with increasing simulated viewing distances. The independent variables were as follows: - Ambient sky luminance (within subject, 2 levels: 0.5 cd/m2, 4 cd/m2) - Pavement type (within subject, 3 categorical levels: New concrete, old concrete, new asphalt) - Pavement marking type (within subject, 3 categorical levels: Patterned tape, flat tape, alkyd paint) - Pavement marking chromaticity (within subject, 8 levels) o Chromaticity 0: (x, y) = (0.430, 0.507) o Chromaticity 1: (x, y) = (0.463, 0.481) o Chromaticity 2: (x, y) = (0.495, 0.455) o Chromaticity 3: (x, y) = (0.390, 0.460) o Chromaticity 4: (x, y) = (0.430, 0.4335) o Chromaticity 5: (x, y) = (0.360, 0.415) o Chromaticity 6: (x, y) = (0.385, 0.400) o Chromaticity 7: (x, y) = (0.340, 0.375) - Eccentricity (between subject, 3 levels: 00, 100, 200). The dependent variables were the response (“yellow” or “white”), response time [sec] (the time it takes for the subject to respond to a particular pavement marking stimulus from the onset), and the location of the transition point (the point where subjects thought that the color of the pavement marking was no longer yellow). The experiment was designed to analyze the effects of pavement surface type, pavement marking type, initial chromaticity of the pavement markings, ambient illuminance as well as the eccentricity (the visual peripheral angle while detecting the pavement marking color) on the

19 judgment of pavement marking color (white vs. yellow) and the time it takes to make that judgment. A computer program was developed to measure and record the response time and the transition point. A total of 42 subjects participated in the experiment. All were above the age of 55. All variables except “eccentricity” were within-subject. Eccentricity was manipulated by shifting the gaze direction (by means of a fixation point on the screen). For the between-subjects variable eccentricity, subjects were divided into three groups, each with 14 subjects (7 males, 7 females). The presentation order of the stimuli was completely randomized and each stimulus condition was repeated for each subject in two replications. Chromaticity in the context of this experiment refers the chromaticity of the closest visible point of the pavement marking at the bottom edge of the viewing screen (see Figure 9). All of the eight initial chromaticity configurations of the presented continuous pavement markings gradually and linearly converged towards the same chromaticity (x, y) = (0.305, 0.321) at the point on the screen that corresponds to 90ft ahead of the vehicle in the real world. Beyond that location, the pavement markings were achromatic. Thus, starting from the closest point (30ft, or 9.14m) up to 90ft (27.4m), the chromaticity of the pavement markings changed at every linear foot, and after 60 iterations at 90ft, the pavement marking was achromatic. 30ft ahead of the vehicle 90ft ahead of the vehicle after which the pavement marking had achromatic appearance 0 degree eccentricity 10 degree eccentricity 20 degreeeccentricity Figure 9. A sample output of POV-Ray with high roadway luminance and high pavement marking luminance at ambient 0.5 cd/m2. The luminance of the pavement markings and the road surface were obtained using the Tarvip model. Thus, depending on the type of pavement marking material, the luminance at a particular point was determined by the retroreflectivity of a previously characterized and modeled yellow material. The selected three materials were yellow patterned tape, yellow flat

20 tape, and yellow alkyd paint. The headlamps were 2000 year model Ford Taurus VOA headlamps in all cases. The selected road surfaces were new concrete, old concrete, and new asphalt. The ambient illuminance (and the luminance of the horizon sky) also had two levels: 0.5 cd/m2 and 4 cd/m2. The luminance of the horizon sky also assumed to affect the luminance of both the road surface and the pavement markings equally, assuming both were of lambertian surface type for the purposes of ambient lighting. The stimuli were generated using POV-ray ray tracer software. A sample stimuli generated with POV-ray is illustrated in Figure 9. Three eccentricity points illustrate the points that subjects were fixated at during the color assessment of the pavement markings. The markings were 4” wide, continuous centerlines 6ft left off the vehicle centerline. The idea was to have different saturations to start with and different gradient cut-offs at the transition point where the color changes from yellow to white along the stripe. With this two-dimensional search space for saturation and distance, we were able to determine the effect of saturation and blending distance on percentage of correct color judgments. The transition of chromaticity from the given eight initial points into the same white point is illustrated in Figure 10. Figure 11 shows a magnified view of the same eight initial chromaticities and their transition path towards white.

21 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 x y Screen gamut Chromaticity-0 Chromaticity-1 Chromaticity-2 Chromaticity-3 Chromaticity-4 Chromaticity-5 Chromaticity-6 Chromaticity-7 (.305, .321) Figure 10. Selected Pavement Marking Color Configurations on CIE 1931 Standard 20 Observer Chromaticity Diagram.

22 0.3 0.4 0.5 0.6 0.2 0.3 0.4 0.5 0.6 x y Chromaticity-0 Chromaticity-1 Chromaticity-2 Chromaticity-3 Chromaticity-4 Chromaticity-5 Chromaticity-6 Chromaticity-7 Spectral Locus Chromaticity-0 (0.430, 0.507) Chromaticity-1 (0.463, 0.481) Chromaticity-2 (0.495, 0.455) Chromaticity-3 (0.390, 0.460) Chromaticity-4 (0.430, 0.4335) Chromaticity-5 (0.360, 0.415) Chromaticity-6 (0.385, 0.400) Chromaticity-7 (0.340, 0.375) All converged to white (0.305, 0.321) Figure 11. Magnified Illustration of Initial and Final Chromaticities Notice that the transition from yellow to white was linear on the CIE diagram for all chromaticities but two, and on the screen 90ft ahead of the observer, all pavement marking chromaticities converged to the point (x, y) = (0.305, 0.321) regardless of the initial chromaticity of the pavement markings. Chromaticity-2 and Chromaticity-4 color configurations did not follow a linear transition because the linear path slightly infringed into the magenta area. RESULTS The dependent variables were investigated to determine the effect of each independent variable and their interactions on these dependent variables. The binary forced-choice responses were analyzed using Generalized Estimating Equations (GEE) as a function of independent variables. GEE features correlated data analysis (repeated measures) methods for binary variables. The number and percentage of yellow and white responses are summarized in Table 12 through Table 19 for chromaticity 0 through chromaticity 7 (Located in APPENDIX C). The GEE analysis revealed that, among the main factors, initial chromaticity of the pavement

23 markings (p<0.01) and horizon sky luminance (p=0.038) were statistically significant in affecting color judgment at α=0.05 significance level. Eccentricity (between-subjects factor), and the roadway surface type failed to reach statistical significance at α=0.05 significance level. Pavement marking type (p=0.077) was just short of having a statistical significance at α=0.05 significance level. The first order interactions between sky luminance and roadway type (p=0.003), pavement marking type and roadway type (p=0.003), chromaticity and roadway type (p≅0.0015), and pavement marking type and chromaticity (p=0.03) were also statistically significant in affecting subjects’ assessment of pavement marking color. Figure 12 shows the selected initial chromaticities and the corresponding percentages of “yellow” responses, and the transition chromaticity curve, overlaid on various color boxes on the CIE 1931 2º standard observer chromaticity diagram. Notice that only P2 was inside the chromaticity limits for nighttime yellow pavement markings, as outlined in the FHWA final rule and ASTM D6628, yielding the highest yellow response rate at 99.3% among the eight selected points. The transition chromaticity curve was generated by connecting the average transition chromaticities, at which subjects would no longer call the color of the continuous pavement marking line “yellow”. The rather odd shape of the curve is due to a systematic response pattern of the subjects: For chromaticities 0, 1, and 2, subjects selected transition points closer to the white point (the left wing of the curve). For less saturated yellow hues administered for chromaticities 5 and 6, the transition points retreated to more saturated yellow (right wing of the curve), toward the points themselves. For the chromaticities in-between, the transition points were also in between. When the initial chromaticity was of a deeper hue, the transition point from yellow to white was closer to white. The collective set of points indicates a general region where the transition from yellow to white occurred. Details of the data are given in APPENDIX C.

24 White P0 , 9 8.5 % P1 , 9 9% P2 , 9 9.3 % P3, 85% P4, 97.4% P5, 29.4% P6, 26.2%P7, 0.9% 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 x y Screen Gamut Transition Curve and Transition Points CIE 1931 Chromaticity Locus OH/DOT Proposed White FHWA Nighttime White FHWA Nighttime Yellow OH/DOT Proposed Yellow Virginia DOT-Drop On Thermo Nighttime FHWA Daytime Yellow FHWA Daytime White Figure 12. Yellow Response Percentages Overlaid on The CIE 1931 Chromaticity Diagram. The other statistically significant main factor was the horizon sky luminance. Note that horizon sky luminance was directly passed on to the pavement marking and road surface luminance. Increasing horizon sky luminance, and corresponding increase in both pavement marking and pavement luminance, from 0.5cd/m2 to 4cd/m2 increased the overall yellow responses from 64% (3899/6048) to 67% (4064/6048).

25 White P0 , 9 8.5 % P1 , 9 9% P2 , 9 9.3 % P3, 85% P4, 97.4% P5, 29.4% P6, 26.2% P7, 0.9% 0.3 0.4 0.5 0.6 0.7 0.3 0.4 0.5 0.6 0.7 x y Screen Gamut Transition Curve and Transition Points CIE 1931 Chromaticity Locus OH/DOT Proposed White FHWA Nighttime White FHWA Nighttime Yellow OH/DOT Proposed Yellow Virginia DOT-Drop On Thermo Nighttime FHWA Daytime Yellow Figure 13. A Magnified View of the Yellow Response Percentages Overlaid on the CIE 1931 Chromaticity Diagram The between-subject variable “eccentricity” did not seem to generate a strong enough difference between yellow and white responses, and therefore is not a statistically significant factor. Hence, between near-foveal, 10º parafoveal, and 20º parafoveal views, the perception of yellow in our case changed only little. Thus, with 95% confidence, we can say that the probability of a subject assessing the same stimulus as yellow for 0º, 10º, and 20º eccentricity levels is essentially the same. The overall number of responses grouped by the “eccentricity” variable is given in Table 3.

26 Table 3. Number of “yellow” and “white” responses for the variable “eccentricity”. Eccentricity [deg] 0 10 20 Yellow 2706 2628 2629 White 1326 1404 1403 N um be r of re sp on se s Total 4032 4032 4032 The road surface type did not affect subjects’ response for pavement marking color strongly enough to generate a statistical significance at α=0.05 significance level. There was a slight increase in overall yellow responses as the roadway surface reflectivity increased (i.e. New concrete), yet again, the difference was not adequately large and consistent to prove statistical significance. Table 4 shows distribution of yellow and white responses grouped by the three road surface types. Table 4. Number of “yellow” and “white” responses for the variable “Road Surface Type”. Road Surface Type New Concrete Old Concrete New Asphalt Yellow 2751 2628 2584 White 1281 1404 1448 N um be r of re sp on se s Total 4032 4032 4032 Pavement marking type was just shy of proving to be a statistically significant factor (p=0.077). Changing pavement marking type affected only the luminance but not the chromaticity. Therefore, pavement marking type refers only to retroreflectivity in the context of this experiment. The general tendency was that the brighter the pavement marking material, the higher the number of white responses, yet this tendency failed to be statistically significant. Table 5 gives the number of responses for each pavement marking category. Table 5. Number of “yellow” and “white” responses for the variable “Pavement Marking”. Pavement Marking Type Patterned Tape Flat Tape Alkyd Paint Yellow 2557 2636 2770 White 1475 1396 1262 N um be r of re sp on se s Total 4032 4032 4032

27 Pairwise comparisons between initial chromaticities showed that the only non- statistically significant differences were found between chromaticity 0, chromaticity 1, and chromaticity 2. All other pairwise combinations of initial chromaticities proved statistically significantly different in affecting the response. The statistically significant first order interactions are detailed in tables in APPENDIX C. Response Time Analysis The response times for all subject responses were measured from the onset of the stimuli to the time that the subjects pressed a mouse button as they made an assessment for the color of the pavement marking stimuli. This was taken as a proxy-measure for the amount of indecision in the subject to choose between yellow or white. Note that the response times were measured regardless of the subject’s color classification. The analysis were again conducted using GEE models, but this time the response variable was on a continuous scale (time in seconds), thus the model family was not binomial but rather Gaussian. The response times were different for different pavement marking initial chromaticities. Road surface type, eccentricity, and horizon sky luminance did not change subjects’ response times enough to prove statistical significance at α=0.05 significance level. Pavement marking type was just short of having a statistical significance (p=0.051) at α=0.05 significance level. Pavement marking chromaticity affected the response time of the subjects. In general, it took longer for the subjects to respond to chromaticities between saturated yellow and white (Chromaticities 3, 4, 5, and 6). For whiter markings (Chromaticity-7), the response time was not as long as those in the confusion area. SUMMARY OF FINDINGS Similar to the results in the color-booth experiment, there was an increase in the “yellow” responses for saturated yellow chromaticities closer to yellow-orange spectra. Overall, however, subjects were more likely to classify a color as yellow when compared to the color-booth experiment. The use of a gradual pavement marking color transition from the starting chromaticity to white may have motivated some of the participants to make comparative color assessments. The use of a white reference color at long distances may have allowed the subjects to compare the two ends of the continuous pavement marking stripe, and respond as “yellow” when they perceived even a slight shade of yellow close in. In contrast, this strategy was not available in the chip-by-chip presentation used in the color booth experiment. As a consequence of this possible relative color judgment strategy, we saw yellow responses for pavement markings with starting chromaticities well into the yellow-green ranges that would have resulted in very low yellow response percentages in the color booth experiment. The reader should also note that in the color booth experiment, “yellow”, “white”, and “neither” were the acceptable responses, but in this experiment “neither” was not an acceptable response.