U.S. Department of Transportation
Federal Highway Administration
1200 New Jersey Avenue, SE
Washington, DC 20590
Federal Highway Administration Research and Technology
Coordinating, Developing, and Delivering Highway Transportation Innovations
|This report is an archived publication and may contain dated technical, contact, and link information|
Publication Number: FHWA-HRT-04-145
Date: December 2005
Enhanced Night Visibility Series, Volume XIV: Phase III—Study 2: Comparison of Near Infrared, Far Infrared, and Halogen Headlamps on Object Detection in Nighttime Rain
PDF Version (929 KB)
PDF files can be viewed with the Acrobat® Reader®
During the preparation for the experimental session, the participants completed a predrive questionnaire, which included a question on the concerns of the drivers when driving at night. This question was used to establish if a driver perceived that he or she was sensitive to glare. None of the younger drivers expressed a concern about glare, but 5 of the 10 older participants named glare or other headlights as an area of concern.
The significant main effects and interactions for each dependent variable are marked with an “x” in table 11. The effect of pedestrian location and its interactions were specific to the disability glare portion of this study.
An ANOVA was performed on the deBoer scale ratings recorded during the driving portion of this study. The model for this portion of the study was a 2 (Adaptation) by 5 (VES) by 3 (Age) mixed-factor design. ANOVA summary tables were developed for the dependent measure of the subjective deBoer scale rating. The complete ANOVA table for the discomfort glare portion appears in appendix I.
Only the main effect of VES glare (VES) was significant for the subjective measure of discomfort (p < 0.05), with an F value of 14.36. The post hoc analysis indicated three significantly different groupings of discomfort glare among the five VESs (figure 10, table 12). The glare produced by the low/narrow halogen VES was rated the most discomforting, with the lowest mean deBoer scale rating of 5.15. This rating of “Just acceptable” was statistically different than the other VESs. On the other hand, the glare produced by the medium/medium HID was rated the least discomforting, with a mean deBoer scale rating of 7.2 (“Satisfactory”). This VES was statistically different from the other VESs, with the exception of the low/wide VES. The high/narrow and high/wide have the same average deBoer rating (6.15).
Figure 10. Bar graph. deBoer discomfort ratings for the main effect of VES (scale of 1 to 9).
An ANOVA was performed on the detection distances taken during the disability glare portion of this study. The model for this variable was a 2 (Driver Light Adaptation Level) by 2 (Pedestrian) by 5 (VES) by 3 (Age) factorial design. ANOVA summary tables were developed for the dependent measurement of detection distance (appendix I). It should be noted that a shorter detection distance suggests more oncoming glare and a longer detection distance suggests less glare. In the graphs for this section, standard error bars are included on top of the means.
A total of 599 observations of the detection distance measurement were gathered during the driving portion of the study, with only one missing datapoint. The results yielded a significant two-way interaction—Pedestrian by VES—and three main effects—VES, pedestrian location, and age.
The main effect of driver light adaptation level was not significant, with an F value of 0.66. The low and high adaptation levels both allowed similar mean detection distances of 94.2 m (309 ft) and 96.6 m (317 ft).
The interaction of pedestrian location and VES was significant (p < 0.05), with an F value of 10.18. As illustrated in figure 11, the low/narrow VES had the lowest detection distance for both the pedestrian on the left and the right. The other VESs had similar distances to each other for the pedestrian on the left; however, the pedestrian on the right appeared to have longer detection for VESs rated as less glaring by the deBoer scale. Low/wide and medium/medium, the two VESs rated the least glaring, allowed detection of the pedestrian on the right at more than 137 m (450 ft). High/narrow and high/wide, the VESs rated as the next least glaring, allowed detection of the pedestrian on the right at approximately 122 m (400 ft). These results indicate that detection of pedestrians on the right may be more susceptible to changes in glare than detection of pedestrians on the left; however, all the pedestrians on the left were detected much later than pedestrians on the right regardless of VES.
Figure 11. Bar graph. Mean detection distances for the interaction of pedestrian and VES.
The main effect of VES was significant (p < 0.05), with an F value of 26.89. The glare produced by the low/narrow halogen headlamps led to the lowest mean detection distance of 66.8 m (219 ft). The post hoc test (figure 12, table 13) showed the same grouping described in the Pedestrian by VES interaction (figure 11). As discussed previously, the grouping of VESs is more likely caused by differences in detection of the right-side pedestrian, with the exception of the poorest-performing VES (low/narrow).
Figure 12. Bar graph. Mean detection distances for the main effect of VES with SNK groupings.
The main effect of pedestrian position was significant (p < 0.05), with an F value of 86.11. The left-pedestrian location yielded a mean detection distance of 67.7 m (222 ft). The mean detection distance for the right pedestrian was much farther at 122.8 m (403 ft).
The main effect of age was significant (p < 0.05), with an F value of 15.92 and had three levels. The post hoc SNK indicated that as age increased, detection distance significantly decreased. Young drivers detected the pedestrians with a mean distance of 114.6 m (376 ft). The mean distance for middle-aged drivers was 95.4 m (313 ft), and for older drivers the distance fell to 76.8 m (252 ft). This trend is illustrated in figure 13.
Figure 13. Bar graph. Mean detection distances for the main effect of age group with SNK groupings.
An ANOVA was performed on the calculated change in illuminance at the driver’s eye ( lux) because of the oncoming glare measured during the disability portion of this study. The specific illuminance levels at the moments of detection were the datapoints of interest; therefore, there was one illuminance reading that coincided with each detection distance. The model for the disability glare portion of this experiment was a
A total of 599 observations of driver’s eye illuminance were gathered during the disability glare portion of the study, with only one missing datapoint. The results yielded a significant three-way interaction: Pedestrian by VES by Age. The two-way interactions of Pedestrian by VES, Pedestrian by Age, and VES by Age were also significant, as well as the main effects of VES, pedestrian, and age. Adaptation level did not result in significant interactions nor a significant main effect. The illumination levels at detection were 1.12 lx and 1.15 lx under the low and high adaptation levels, respectively.
The interaction of pedestrian and VES and driver age was significant (p < 0.05), with an F value of 7.75. The primary cause of the interaction was the mean illuminance of 5.69 lx at the moment of detection for older drivers viewing low/wide headlamps with the left pedestrian, whereas the same scenario produced an illuminance level of 0.90 lx at the moment of detection for younger drivers (figure 14). A further analysis of this data indicated that exceptionally high illuminance values were recorded for 7 of the 10 older participants at the time of detection. In an effort to isolate the cause of these high illuminance values, the data were reviewed in more detail. It appeared that the high values all occurred during the end of the data collection effort. In addition to these seven older participants, two middle-aged participants and one younger participant also participated during this time period. These three participants also had high illuminance values, indicating that something may have occurred to the headlamp during this time period that was not detected by the experimental team. Even with this possible confound, the low/wide VES was the second least glaring and allowed the longest detection distance of the pedestrians; however, if the last 10 participants had experienced the same glare level as the first 20 participants for this VES, this VES may have been rated as less glaring and allowed greater detection distance. More detail can be found in chapter 4, Discussion.
Figure 14. Bar graph. Mean illuminance readings (lx) at moment of detection for the Pedestrian by
The interaction of VES by Age was significant (p < 0.05), with an F value of 5.7. The primary reason this interaction was significant is because of the older participants with the low/wide VES as discussed above. Other than this effect, it appears that the high/narrow VES and the low/narrow VES follow the expected trend of younger participants experiencing the least illuminance, older participants experiencing the most illuminance, and middle-aged participants being in between the extremes (figure 15). On the other hand, the medium/medium VES had similar illuminance for both the middle and older age groups, and the high/wide VES had similar illuminance for both the younger and middle age groups.
Figure 15. Bar graph. Mean illuminance readings (lx) at moment of detection for the VES by Age interaction.
The interaction of Pedestrian by Age was significant (p < 0.05), with an F value of 6.0. The primary reason this interaction was significant is also because of the older participants with the low/wide VES as discussed in the VES by Age by Pedestrian interaction paragraph. The left pedestrian was the pedestrian associated with the high illuminance for the low/wide headlamp, causing the interaction shown in figure 16.
Figure 16. Bar graph. Mean illuminance readings (lx) at moment of detection for the Pedestrian by Age interaction.
The interaction of Pedestrian by VES was significant (p < 0.05), with an F value of 12.4. Although this interaction is also influenced by the Pedestrian by Age by VES interaction, there are some other interesting aspects of this interaction. As shown in figure 17, both of the wide-beam VESs had the largest illuminance at detection of the left pedestrian, which was also more than three times that of their illuminance at detection of the right pedestrian. The medium/medium VES also seemed to follow this trend; however, both the narrow-beam VESs had similar illuminances at the point of detection for both the left and right pedestrians.
Figure 17. Bar graph. Mean illuminance readings (lx) at moment of detection for the Pedestrian by VES interaction.
The main effect of VES included five different sets of VESs. The SNK revealed three significantly different groupings of driver’s eye illuminance ( lux) among the five VESs (figure 18, table 14). The low/narrow VES produced the highest mean driver’s eye illuminance level at detection, 1.93 lx. Both the wide-beam VESs were in the second group with similar illuminances at the point of the detection. The medium/medium VES and high/narrow VES were grouped together with the lowest illuminance at detection (0.46 lx and 0.54 lx, respectively).
Figure 18. Bar graph. Mean illuminance readings (lx) at moment of detection for the main effect of VES with SNK groupings.
The main effect of pedestrian included the left- and right-pedestrian locations. The left-pedestrian location had a mean illumination level at detection of 1.52 lx. The mean illumination level for the right pedestrian was approximately half as high, 0.75 lx.
The main effect of age indicated that as age increased, illumination at detection increased (figure 19). Young drivers had a mean illumination level at detection of 0.77 lx. The mean illumination level for middle-aged drivers was 0.93 lx, and for older drivers the level rose to 1.72 lx. The younger and middle-aged participants were not statistically different from each other, but both age groups had significantly less illuminance at detection than the older age group.
Figure 19. Bar graph. Mean illuminance readings (lx) at moment of detection for the main effect of
age group with SNK grouping.