Information As A Source of Distraction

Chapter 2. Laboratory Evaluation of CMS

Introduction

Laboratory comparisons were made between messages displayed on a CMS and the same messages displayed on a 60-inch (1.5-m) LED/LCD monitor. The purpose of the comparisons was to determine which display properties of CMS messages were most important when emulating messages on other types of displays. There were two immediate applications for the findings: (1)to guide the display of CMS messages in the FHWA sign laboratory, which uses the LED/LCD display used in these comparisons, and (2) to guide the display of CMS messages in the FHWA highway driving simulator.

The FHWA Human Factors Team often evaluates new sign formats in its sign laboratory. In the past, these formats have been for static highway signs. In those evaluations, red, green, and blue color values and visual angle subtended were the primary concerns in emulating the highway signs. This seemed reasonable given that in the real world, the actual color values of the signs would vary with the available lighting, and visual angle would vary with viewer distance. However, modern CMSs use LEDs to emit light so available light will primarily affect contrast and not color. The LEDs on the Daktronics® CMS used in this study consist of one red, one blue, and one green LED in each pixel, and the pixels are spaced 0.79 inches (20 mm) apart, center to center. At sufficiently long distances, the light emitted from individual LEDs is perceived as coming from a single source, and the individual pixels are perceived to blend together, much as individual color television pixels, when viewed from an appropriate distance, are not perceived. The present study examined whether it would be necessary to emulate the individual LED elements or pixels of the CMS when presenting alternative CMS message formats in the sign laboratory or the driving simulator.

Method

Materials

The CMS was a Daktronics® VF-2320 full-color (red, green, blue (RGB)) matrix display with 0.79-inch (20-mm) pixel pitch. The display surface was 4 by 4 ft (1.2 by 1.2 m), which accommodated a 64- by 64‑pixel display. The display is shown in figure 19 in chapter 3.

Some messages for the CMS were developed using the Daktronics® Vanguard V4 Professional software. Other messages for the CMS were images of static signs adapted from the MUTCD and supplied by Daktronics®. Bitmaps were exported from the Daktronics® software and then imported into graphics software to develop messages to be displayed on the LCD/LED monitor.

The graphics software was used to modify the imported bitmaps so that alternative methods of emulating the CMS display could be compared. In some cases, an image was intended to come as close as possible to a literal replication of the CMS display—where each pixel consisted of three circular RGB elements and pixels were 0.79 inches (20 mm) apart, center-to-center. In other cases, the emulated images were made up of circular or rectangular pixels of a uniform color. Finally, some images were made of filled in tracings of the imported images—that is, individual pixels were not emulated in those images.

The 60-inch (152-cm) LED/LCD display was shorter and wider (29 by 52 inches (73.6 by 132cm)) than the CMS. To display images on the CMS that were the same size as those on the LCD, the CMS images were cropped to the LCD height.

Procedure

Display comparisons were made in the FHWA photometric and visibility laboratory, which has flat black interior walls and no windows and is sealed so that outside light is not admitted. During the comparisons, the only light inside the laboratory came from the two display devices, the CMS controller, and a laptop computer used to control the LCD.

The intensity of the CMS was set to the lowest available setting (1 percent). At that setting, a white circular area with a 12-inch (30.5-cm) diameter emitted about 52.5 fl (180 cd/m²⁾. A circular area consisting of only red LEDs emitted about 14.6 fl (50 cd/m²⁾, a blue circular area emitted about 4.4 fl (15 cd/m²⁾, and a green circular area emitted about 36.5 fl (125 cd/m²⁾. Table 1 shows the photometric measurements of the X and Y Commission Internationale de l'Eclairage (CIE) color coordinates of stimuli displayed on the CMS.⁽⁴⁷⁾ The table also shows the specified RGB settings that were used to generate the measured colors.

Table 1. Measured CIE color coordinates for colors displayed on the CMS and the RGB settings used to generate those colors.

On the LCD, a white circular area with a 12-inch (30.5-cm) diameter measured 32.4 fl (111 cd/m²⁾; a red circular area emitted 7.4 fl (25.2 cd/m²⁾, green emitted 23.5 fl (80.4 cd/m²⁾, and blue emitted 2.3 fl (8.0 cd/m²⁾. Table 2 shows the photometric measurements of the X and Y CIE color coordinates of colors displayed on the LED/LCD monitor.

Table 2. Measured CIE color coordinates for colors displayed on the LED/LCD and the RGB settings used to generate those colors.

For each pair of stimuli presented, one stimulus on the LED/LCD and one on the CMS, participants were instructed to rate similarity, where a rating of 1 indicated that the pair was not at all similar and a rating of 7 indicated that the pair was very similar.

Participants first compared the images on the two displays from a distance of 90 ft (27 m) with the LCD immediately to the left (relative to the observer) of the CMS display. The images on both displays were at the same height above the floor, and subtended the same visual angle.

The comparisons were then repeated with the LCD positioned 15 ft (4.6 m) from the participant and slightly to the participant’s left so that the LCD and the CMS, which was still 90 ft (27 m) away, could be viewed at the same time. The images presented at the near distance were sized so that they subtended the same visual angle as the images on the CMS. Except for the reduced size of the near images, they were as similar as possible to the CMS images.

Participants

Nine individuals—FHWA Federal employees and contract staff—served as subjects.

Stimuli

To ease the burden on the reader in relating stimuli to results, the individual stimulus pairs are described in the Results section along with the comparison findings.

Results

The results for the comparisons with the two displays side by side are presented first. These are followed by the same signs with the LCD at 15 ft (4.6 m).

Exit Closed (Side-by-Side)

There were four versions of the sign shown in figure 3. They varied only in RGB values, which are shown in table 3. There were no significant differences in the similarity ratings between versions.

Figure 3. Photo. Message advising that exit is closed.

Table 3. RGB values used in comparisons with message advising that exit is closed.

Exit Closed (LCD Close, CMS Far)

As with the side-by-side comparison, there were four versions of the close comparison stimuli. Unlike the side-by-side stimuli, there was no attempt to emulate individual pixels; all characters were solid. An example of the stimuli as they appeared on the LCD screen is shown in figure 4.

Figure 4. Photo. Exit closed sign example (proportionally scaled and zoomed in).

There were no significant differences in ratings among the four comparisons, which used the same color settings as shown in table 3. Across the four comparison stimuli, the overall mean rating was 4.9.

End (Side-by-Side)

The two versions of this sign are shown in figure 5 and figure 6. Version A consists of white and grey circles intended to simulate the appearance of individual pixels. Version B consists of red, green, and blue circles intended to simulate the three LEDs that make up individual pixels. Both images were based on the same bitmap supplied with the sign vendor’s software.

Figure 5. Photo. Version A of the end sign.

Figure 6. Photo. Version B of the end sign.

Version A with a mean similarity rating of 5.0 (with a standard error of the mean (se) = 0.236) was rated significantly more similar to the CMS display of the bitmap than was version B, which was rated 4.2 (se = 0.324), F(1, 8) = 12.25, p = 0.008. This suggests that at a simulated distance of 90 ft (27 m), it is not necessary to simulate individual LEDs. Use of red, green, and blue circles considerably reduced the perceived brightness of the LCD image, which made the image appear grey in comparison with the nominally white CMS image.

End (LCD Close, CMS Far)

There were three stimuli for this comparison. Two were smaller versions of those shown in figure 5 and figure 6. The additional comparison stimulus eliminated the space between pixels and is shown in figure 7.

Figure 7. Photo. Solid version end sign comparison stimulus.

There were no significant differences in the ratings of the three comparison stimuli. The overall stimuli mean rating was 4.8.

High Wind (Side-by-Side)

There were four versions of the high wind warning sign, one of which is shown in figure 8. Three of the four signs displayed on the monitor were rated reasonably similar to the same sign on the CMS. The difference in mean rating among the four versions was significant, F(3, 24,) = 8.67, p = 0.012. Post hoc tests showed that version C which had a more orange appearance, was rated lower than the other versions, F(1, 8) = 64.0, p < 0.0001. The mean and 95-percent confidence limits for the similarity ratings are shown in table 4.

Figure 8. Photo. Emulation of high wind warning sign.

Table 4. RGB values and similarity ratings of four versions of the high-wind warning sign with the CMS standard.

High Wind (LCD Close, CMS Far)

In reducing the size of the high wind stimuli in which individual pixels were simulated, the brightness was reduced to the point where the LCD stimuli were barely visible. Therefore, the space between pixels was dropped. Version B of the comparison stimuli, as described in table 5, is shown in figure 9. The differences between ratings were significant, by Wilk’s Lambda, F (3, 6) = 5.7, p = 0.034. Versions B and D were rated significantly higher than A and C.

Table 5. RGB values and rating results for the close high wind comparison stimuli.

Figure 9. Photo. Version B of the high wind comparison stimuli.

Accident Ahead Merge Left (White, Side-by-Side)

There were three versions of the accident ahead text message that were nominally white. These are shown in figure 10 through figure 12. One version consisted of white pixels (red, green, and blue for each set to 255), the second grey pixels (red, green, blue for each set to 230), and the third in which the space between pixels was filled in with white (red, green, and blue for each set to 255). There was a significant difference in similarity ratings among the three signs, F(2,16)=7.3, p=0.006. Post hoc comparisons showed that the sign with grey pixels was rated significantly less similar to the CMS image than the two white versions. At a distance, the space between pixels, whether black or white, resulted in images that were judged equally similar. Comparisons were not requested with the close LCD because the small pixelated images appeared too dim to warrant testing.

Figure 10. Photo. Accident ahead merge left sign with white pixels that yielded a mean similarity rating of 5.6 between the CMS and LCD displays.

Figure 11. Photo. Accident ahead merge left sign with grey pixels that yielded a mean similarity rating of 4.7 between the CMS and LCD displays.

Figure 12. Photo. Accident ahead merge left sign with solid white letters that yielded a mean similarity rating of 5.6 between the CMS and the LCD displays.

Accident Ahead Merge Left (Orange)

Four versions of the accident ahead merge left message in various shades of orange or amber were used, one example of which is shown in figure 13. An overall test indicated significant differences in mean similarity ratings, F(3, 24) = 15.1, p < 0.001. Post hoc tests showed that versions C and D in table 6 were rated significantly more similar to the CMS image than versionA. Version B was not significantly different from version A. Interestingly, version B had the same RGB values as the standard displayed on the CMS. It appears that in this case, the Samsung® display required a greater amount of green to approach the perceived color on theCMS.

Figure 13. Photo. Accident ahead merge left example.

Table 6. RGB values for the four orange alternatives to the accident ahead merge left standard.

No Pedestrians

There were two LCD versions of the no pedestrians (no peds) text message, which are shown in figure 14 and figure 15. One of the versions was made up of individual white dots aligned in the same manner as the pixels on the CMS. The other version was similar, but the spaces between pixels were filled in so that only the outer edges of characters retained the outline of the CMS pixels. Both versions received equal and relatively low similarity ratings (Mean (M) = 4.8).

Figure 14. Photo. No peds message on LCD with emulated CMS pixels.

Figure 15. Photo. No peds message on LCD with filled-in space between emulated CMS pixels.

Exit 47 B, Route 7 N, Closed

Similar to the no pedestrians sign, the Exit 47 B sign comparison was a pixelated emulation, shown in figure 16, and an emulation with solid letters, shown in figure 17, of the same message on the CMS display. The solid version received a slightly higher rating (5.3) than the pixelated version (5.0), although this difference did not reach statistical significance, F(1, 8) = 4.0, p = 0.08.

Figure 16. Photo. Exit 47 B message on LCD with emulated CMS pixels.

Figure 17. Photo. Exit 47 B message on LCD with filled-in space between emulated CMS pixels.

Shoulder Work

Four versions of a work zone warning (see figure 18) with negative contrast were compared.
The RGB values of the LCD images and the CMS image are shown in table 7. There were no significant differences among the four ratings. It appears that the reasonable variation in orange shading of the simulated signs is not critical.

Figure 18. Photo. Shoulder work CMS message emulation on LCD.

Table 7. Ratings for negative contrast orange shoulder work sign.

Discussion

The ratings suggest that few participants would mistake an LCD image for an image on the CMS. Only one participant gave any ratings of 7. On the other hand, many of the LCD images received mean ratings of 5 or more, which may suggest that the LCD may be a useful tool in addressing human factors issues with CMS messaging.

Overall, the findings suggest that when simulating a CMS display, attempts to emulate individual pixels are counterproductive. This makes sense given that even at its lowest brightness setting, the CMS is brighter than the LCD used in these tests. Emulating CMS pixels reduces the brightness of the LCD even more than when all LCD pixels are used.

Subjective color matches for LCD stimuli are preferable to attempts to match the RGB settings of the CMS. The preference for subjective color matches can be seen in several of the results. For instance, the LCD accident ahead message that was judged most similar to the CMS message had more green in the RGB setting than did the CMS. This also makes sense given the brightness of the LCD color boxes compared with those of the CMS.

When emulating CMS displays, whether in the laboratory or on projected images in a driving simulator, the intended appearance of the CMS should be the primary concern not the absolute color settings on the CMS or the pixel content of the CMS display.