Evaluation of The Impact of Spectral Power Distribution on Driver Performance

CHAPTER 4. Scoping Experiment

Introduction

This experiment was the initial investigation of the spectral impact of overhead lighting and vehicle headlamps on driver visual performance. It was a scoping experiment that provided the framework for subsequent investigations.

Research Objectives

The purpose of the scoping experiment was to accomplish the following tasks:

• Evaluate the effect of the spectral distribution of overhead-lighting sources on driver ability to detect pedestrians and targets and recognize colors in the environment.
• Evaluate the effect of the spectral distribution of vehicle headlamp color on driver ability to detect pedestrians and targets and recognize colors in the environment.
• Evaluate the impact of the overhead-lighting color and headlamp color in mesopic conditions on driver ability to detect pedestrians and targets and recognize colors in the environment.
• Evaluate the impact of spectral distribution of overhead lighting and headlamp lighting on detection of pedestrians and targets located peripherally.
• Provide a framework upon which to design subsequent experiments.

Experimental Design

A 2 by 4 by 2 by 2 by 4 mixed-factors experiment was designed to investigate the relationship of the spectral distribution and level of overhead lighting and headlamps on object detection and recognition on a roadway. A partial factorial design was used to ensure participants of both age groups were exposed to all combinations of vehicle speed, overhead lighting, and headlamp colors and intensities. Experimental variables are listed in table 6 and table 7.

Table 6. Scoping experiment independent variables and values.

Table 7. Scoping experiment dependent variables and measurement method.

Independent Variables

Age

Age negatively affects visual perception.⁽⁴⁸⁾ There are two components to this visual reduction. The first is the yellowing of the lens as a person ages, which acts as a filter across the eye, and the second is intraocular scatter, in which light is significantly scattered as it passes through the ocular media thus reducing the contrast and the ability of the driver to detect objects. (See reference 49, 51, 57, and 94.) Burton, Owsley, and Sloane found that the reduction in retinal image quality is almost linear with the age of the observer.⁽⁹⁴⁾ However, it is noteworthy that driver performance may not follow the same trend as the reduction in the retinal image because older drivers have more experience driving in general and at night.

To examine the effects of age on visual performance, two age groups were investigated: younger drivers (25–35 years old) and older drivers (65 years old and older).

Overhead-Lighting Type and Level

For this experiment, the 2,100-K HPS and the 6,000-K LED overhead-lighting systems were used. Two lighting levels were used: 1.25 lx and 5 lx (0.12 and 0.46 fc).

Headlamp Configuration

Four headlamp configurations were used in this experiment: white/yellow low intensity, white/yellow high intensity, white/blue low intensity, and white/blue high intensity. These colors were chosen to match the color temperature of the HPS overhead lighting and LED vehicle headlamps, respectively. Details regarding the headlamps and filters are included in chapter 3, Experimental Approach.

Pedestrian Clothing Color

All four pedestrian clothing colors were used in this experiment.

Pedestrian Position

During the experiment, the confederate pedestrians stood in three sets of positions to satisfy threeexperimental needs.

The first set of positions was designed to isolate the effect of roadway lighting type and level. The VI falling on the pedestrians was controlled. The pedestrians were placed along the length of the Smart Road so that the roadway lighting types (HPS and LED) would project the same VI on the pedestrian across both lighting types and levels. Depending on the lighting type and level, each pedestrian would have to move to one of four possible locations along the road to achieve the same VI. In each location, the pedestrian was 0.6 m (2 ft) to the right of the right shoulder line, shown in figure 35. This was called the constant-VI position.

The second set of positions (the constant-contrast positions) was designed to keep the pedestrian’s contrast with the background constant. For the constant-contrast position, the pedestrian stayed in the same place, 0.6 m (2 ft) from the right shoulder line, for both intensity levels of the same lighting type. This means that when the lighting system dimmed, the VI would change but the object contrast would not because the background would also be dimmed.

Figure 35. Photo. Pedestrian on roadway.

The last set of positions was designed to investigate off-axis visibility and to represent a possible collision hazard instead of an immediate collision hazard. This pedestrian was placed beyond the right guardrail, approximately 18 m (59 ft) off the roadway. That was called the off-axis position.

The design was balanced so that all clothing colors and pedestrian positions were viewed under all the overhead lighting and vehicle headlamp lighting combinations.

In the two positions close to the road, pedestrians stepped away from the road when the participant identified them or when participant vehicle reached a certain point.

Target Color and Position

All four colors of targets were used in this experiment. During the experiment, the targets were placed 0.6 m (2 ft) to the right of the shoulder with the flat face aimed toward the oncoming participant driver. The target positions controlled for VI, similar to the constant-VI position of the pedestrians. Figure 36 is a photo of the target in the roadway.

Figure 36. Photo. Target on roadway.

Speed

The experiment was performed at 89 and 56 km/h (55 and 35 mi/h).

Dependent Variables

Detection and color-recognition distances were measured in this experiment, as described in chapter 3. Orientation-recognition distance was not measured, and pedestrians and targets faced the same direction for all runs.

Methods

Facilities and Equipment

The experiment was performed on the Smart Road using the equipment described in chapter 3, Experimental Approach.

Participants

The participants were recruited and screened, as described in chapter 3. Thirty-two participants completed the experiment and were balanced by both age and gender. Mean and standard deviation of participant age, visual acuity, mesopic visual acuity, low contrast visual acuity, and UFOV are listed in table 8.

Table 8. Scoping experiment participant characteristics.

Procedure

Experimental Sessions

Those eligible for the experiment, based on the vision and health screening, were asked to come to two experimental sessions, with each session testing one overhead-lighting type and speed. Two participants completed the experiment during each session.

Data Collection

Once on the Smart Road, participants drove one practice lap followed by eight experimental laps. During the experiment, on-road experimenters altered the overhead-lighting conditions, headlamp filters, object types, colors, and positions according to the protocol. In-vehicle experimenters directed the participant to drive at the correct speed and recorded detection and color-recognition distances using button presses.

Data Analysis

After data reduction, dependent variables were analyzed with respect to the independent variables. If the participant failed to detect a pedestrian or target, it was counted as a miss. If the participant recognized the incorrect color, it was counted as a miss.

The detection and color-recognition distances for pedestrians and targets were compared across the experimental conditions using analysis of covariance (ANCOVA) in Statistical Analysis System® software. Individual and interaction effects were analyzed. Independent variables were compared within three groups: on-axis pedestrians, off-axis pedestrians, and targets. There were too few detections of off-axis pedestrians to draw meaningful conclusions; however, results are described here. Results for on- and off-axis pedestrians were also compared.

Student-Newman-Keuls (SNK) post hoc tests were performed for all significant effects to determine the contribution of the individual factors to the statistical significance. In all figures reporting SNK results, mean values sharing a letter label are not significantly different from each other.

Results

Overhead-Lighting Type and Level

For runs in overhead lighting, independent variables were analyzed with respect to on- and off-axis pedestrian and target detection and color-recognition distances. Results for on- and off-axis pedestrians were also compared.

On-Axis Pedestrian

ANCOVA results are listed in table 9, with significant main effects described below the table.

Table 9. Scoping experiment on-axis pedestrian significant results summary for runs in overhead-lighting type and level.

Age:

Driver age significantly affected pedestrian-detection distance. Younger drivers detected pedestrians from significantly farther away (the mean M = 118 m (386 ft)) than older drivers (M= 88.5 m (290 ft), p = 0.0092). Age significantly affected color-recognition distance for pedestrian clothing. Younger participants recognized the clothing color from significantly farther away (M = 69.3 m (227 ft)) than older participants (M = 49.1 m (161 ft), p = 0.0108). Both effects were expected because previous research has shown these types of effects. These may be a result of intraocular scatter or the natural yellowing of the lens as a person ages.

Overhead Lighting:

Overhead-lighting type and level significantly affected detection distances and color-recognition distances. Higher lighting levels of both types corresponded to longer detection distances, as seen in figure 37, where bars sharing a letter are not significantly different from each other. What is important to note is the difference within overhead-lighting type. The difference in detection distances between the high and low levels for the 6,000-K LED luminaires was statistically significantly greater than the difference between the high and low levels for 2,100-K HPS (p = 0.0023) sources. This indicates an impact that can be attributed to the spectral distribution of the light source.

Overhead-lighting type and level affected pedestrian color-recognition distance similarly to how they affected pedestrian-detection distance—detection distances were significantly longer for higher level lighting. Again, the difference between the high and low levels was more pronounced for the LED lighting than for the HPS lighting, again showing an impact owing to the overhead lighting’s spectral distribution. The color-recognition distance results are shown in figure 38, where bars sharing a letter are not significantly different from each other.

1 m = 3.3 ft

Figure 37. Chart. Scoping experiment—mean pedestrian-detection distance by overhead-lighting type and level, with SNK results.

1 m = 3.3 ft

Figure 38. Chart. Scoping experiment—mean pedestrian color-recognition distance by overhead-lighting type and level, with SNK results.

Pedestrian Clothing Color:

Pedestrian clothing color significantly affected detection distances. Figure 39 shows that when the four lighting conditions were compared, gray clothing was detected from significantly farther away than any other color, and red was detected from significantly closer than any other color. Bars sharing a letter are not statistically significantly different from each other.

Pedestrian clothing color also significantly affected color-recognition distance but in a different manner than detection distance. The pedestrian with blue clothing was detected, on average, at significantly shorter distances than the pedestrians wearing the other colors see (figure 40). Bars sharing a letter are not significantly different from each other.

1 m = 3.3 ft

Figure 39. Chart. Scoping experiment—mean pedestrian-detection distance by clothing color.

1 m = 3.3 ft

Figure 40. Chart. Scoping experiment—mean pedestrian clothing color-recognition distance by clothing color.

Off-Axis Pedestrian

Participants detected 23 percent of off-axis pedestrians. Participants had only five opportunities to detect off-axis pedestrians during the course of the experiment, too few to draw strong conclusions, and no effects were found to be significant. However, those results provided direction to the research team for future experiments and are discussed.

Overhead-Lighting Type and Level:

Off-axis pedestrians were detected from farther away, on average, in LED overhead lighting (M = 55.7 m (183 ft)) than in HPS overhead lighting (M = 44.7 m (147 ft)). The difference was not statistically significant, but it is in the expected direction, and overhead lighting is capable of off-axis illumination. In addition, when overhead lighting was broken down by color and level, off-axis pedestrians were detected from farthest away in high-level LED lighting, followed by low-level LED lighting. Those results show promise that LED lighting, efficient in the mesopic range, allowed objects on the road’s periphery to be detected from farther away, more so than HPS lighting at higher levels. A comparison between on- and off-axis pedestrian-detection distances in the two lighting types is shown in figure 41.

1 m = 3.3 ft

Figure 41. Chart. Scoping experiment—mean off-axis pedestrian-detection distance by overhead-lighting type and level.

When overhead-lighting type and level were combined, they had a significant effect on off-axis pedestrian color-recognition distance. Participants recognized the color of off-axis pedestrian clothing color from farther away in LED lighting (M = 24.2 m (79.4 ft)), than in HPS lighting (M= 19.8 m (64.9 ft), p > 0.05. The results, shown in figure 42, were in the expected direction given off-axis spectral effects in the mesopic range but were not significant.

1 m = 3.3 ft

Figure 42. Chart. Scoping experiment—mean off-axis pedestrian color-recognition distance by overhead-lighting type and pedestrian position.

On- Versus Off-Axis Pedestrian

Overhead-Lighting Type:

Overhead-lighting type did not significantly affect color-recognition distance for either on- and off-axis pedestrians, but in both cases, the mean color-recognition distances for LED lighting were greater than for HPS. The effect was more so for the off-axis pedestrian, visible in figure 43. This could indicate a spectral effect in which bluer LED lighting, which is more efficient for off-axis mesopic vision than HPS lighting, led to better pedestrian detection. Subsequent experiments in this project examined off-axis object detection in detail.

Clothing Color:

For on-axis pedestrians, blue clothing had a much shorter color-recognition distance than clothing of the other colors, but the same trend was not seen for off-axis pedestrians, nor for the detection distance of this blue-clothed pedestrian (figure 44). Theoretical explanations regarding human vision do not explain those findings. One environmental factor could be that the on-axis pedestrians were seen against the road, and the off-axis pedestrians were seen against the hills and brush on the roadside. It could be that the differently contrasting environments affected color recognition differently and unpredictably.

1 m = 3.3 ft

Figure 43. Chart. Scoping experiment—pedestrian-detection distance by position and clothing color.

1 m = 3.3 ft

Figure 44. Chart. Scoping experiment—pedestrian color-recognition distance by position and clothing color.

Speed and Overhead-Lighting Type:

Overhead-lighting levels were pooled for this analysis. For on-axis pedestrians, mean detection distances at 89 km/h (55 mi/h) were shorter than at 56km/h (35 mi/h), but the opposite trend, visible in figure 45, was seen for off-axis pedestrians. The detection distance for 56 km/h (35 mi/h) was particularly short because detection distances for HPS were short. However, with so few detections of off-axis pedestrians, it is difficult to draw a firm conclusion that the shorter distance is due to the overhead-lighting type.

1 m = 3.3 ft

Figure 45. Chart. Scoping experiment—mean pedestrian-detection distance by pedestrian position, speed, and overhead-lighting type.

Constant Contrast Pedestrian

An analysis of the constant-contrast pedestrian found a significant effect of participant age on detection and color-recognition distances. Results for all factors are listed in table 10.

Table 10. Scoping experiment constant-contrast pedestrian results summary.

Overhead-lighting level and type did not significantly affect detection distance for the constant-contrast pedestrians. The constant-contrast pedestrian was in the same location for each detection, with the same background, regardless of overhead-lighting level. Lower lighting levels had slightly longer detection distances for both overhead-lighting types, as shown in figure 46, but the effect was not statistically significant.

1 m = 3.3 ft

Figure 46. Chart. Scoping experiment—detection distances by overhead-lighting type and level for constant-contrast pedestrians.

Targets

The significant effects of the independant variables on detection and color-recognition distances of the targets for detections performed in overhead lighting are summarized in table 11.

Table 11. Scoping experiment target results summary for overhead lighting.

Target Color:

Only target color significantly affected target detection distances in all overhead-lighting conditions. When the four overhead-lighting conditions were combined, the blue target was detected from significantly farther away than the other targets, and the gray target was detected from significantly closer (figure 47), where bars sharing a letter do not statistically differ from each other. The road targets were detected on-axis, and color would not necessarily be prone to mesopic effects. More experiments were performed during this project to examine spectral effects in the mesopic range.

1 m = 3.3 ft

Figure 47. Chart. Scoping experiment—mean target-detection distance by target color.

Target color significantly affected target color-recognition distance. Red was detected from the farthest away, followed by green, blue, and gray (figure 48). All color-recognition distances differed significantly from each other. The color-recognitions distances (red, green, blue, and gray) are in a different order than the detection distances (blue, red, green, and gray). That could be because of confusion between blue and green. Color contrast between the target and the background could have an effect, but the luminance camera did not take color-contrast data, so color contrast could not be analyzed.

1 m = 3.3 ft

Figure 48. Chart. Scoping experiment—mean target color-recognition distance by target color.

Target Color and Overhead-Lighting Type:

Target color and overhead-lighting type combined to affect target color-recognition distance. Overhead-lighting levels were pooled for this analysis. In LED lighting, the red and green targets were visible from a greater distance than in HPS lighting. That effect was not seen for the blue and gray targets, as shown in figure 49.

1 m = 3.3 ft

Figure 49. Chart. Scoping experiment—target color-recognition distance by target color and overhead-lighting type.

No Overhead Lighting

For experiments without overhead lighting, independent variables were analyzed with respect to pedestrian and target detection and color-recognition distances, and results are shown in table 12 for pedestrians and table 13 for targets. Results for off-axis pedestrians were not analyzed for runs without overhead lighting because there were too few detections for results to be meaningful.

Pedestrian

Table 12 summarizes scoping experiment results for pedestrians with no overhead lighting.

Table 12. Scoping experiment pedestrian results summary for no overhead lighting.

Pedestrian Clothing Color: Pedestrian clothing color significantly affected detection distance, with gray clothing being detected from significantly farther away than the other clothing colors (p < 0.0001) (figure 50). As shown in figure 50, the red and gray clothing colors had higher reflectances than the blue and black. When illuminated with vehicle headlamps alone, the higher-reflectance pedestrians were more positively contrasted and more visible.

1 m = 3.3 ft

Figure 50. Chart. Scoping experiment—mean pedestrian-detection distance by clothing color for no overhead lighting.

Pedestrian clothing color significantly affected pedestrian clothing color-recognition distance (p < 0.0001) (figure 51). Pedestrians with red and gray clothing were recognized from farther away than with black and blue clothing. The similar and shorter color-recognition distances for black and blue might be because participants found it difficult to differentiate between the two dark colors. Target detection followed the same pattern.

1 m = 3.3 ft

Figure 51. Chart. Scoping experiment—mean pedestrian color-recognition distance by clothing color for no overhead lighting.

Targets

Table 13 summarizes scoping experiment results for targets with no overhead lighting.

Table 13. Scoping experiment target results summary for no overhead lighting.

Target Color:

Target color significantly affected target detection distance (p = 0.0242), with gray targets detected from significantly farther away than green targets, but no other pairwise comparisons were significant (figure 52).

1 m = 3.3 ft

Figure 52. Chart. Scoping experiment—mean target-detection distance by target color for no overhead lighting.

Target color significantly affected color-recognition distance, with red and green target colors recognized from significantly farther away (red M = 37.3 m (122 ft); green M = 35.6 m (17 ft)) than blue targets (M = 28.8 m (94.4 ft), p = 0.0312) (figure 53).

1 m = 3.3 ft

Figure 53. Chart. Scoping experiment—mean target color-recognition distance by target color for no overhead lighting.

Luminance Analysis

Luminance camera images taken at the moment a participant detected a pedestrian or target were analyzed. They were also analyzed at color recognition, but those results largely mirrored the detection results and are not discussed. Luminance analyses were only performed for runs with overhead lighting.

Pedestrian-Detection Distance

Luminance:

At the moment participants detected pedestrians under HPS lighting, the average pedestrian luminance was 0.327 cd/m² (0.095 fL), less than that in LED lighting, 0.464 cd/m² (0.14 fL). Figure 54 shows a threshold at a luminance of approximately 0.2 cd/m² (0.06 fL) where on-axis pedestrians, regardless of clothing color, were able to be detected.

1 cd/m² = 0.3 fL
1 m = 3.3 ft

Figure 54. Scatter Plot. Scoping experiment—pedestrian-detection distance, luminance, and clothing color.

Contrast:

The average contrast when pedestrians were detected was also different between the lighting types, with the contrast in HPS lighting at -0.086 and LED lighting at -0.032. A lower negative contrast, as seen in HPS lighting, means a darker pedestrian with respect to the background was needed than in LED lighting.

When further broken down by clothing color, a combined effect of clothing color and lighting type on contrast at time of detection is seen in figure 55. For all clothing colors except red, pedestrians were more highly contrasted in LED lighting than in HPS lighting, as measured by detection distance. In HPS lighting, red clothing requires the lowest contrast for detection, meaning it is particularly visible in that lighting type. The opposite trend is seen in LED lighting, where red requires the greatest contrast for detection, meaning it is less visible in that lighting type. That is because the spectrum of HPS lighting is redder than that of LED lighting, making red clothing visible, while the opposite is true in LED lighting.

Figure 55. Chart. Scoping experiment—pedestrian mean Weber contrast at time of detection by overhead-lighting color and clothing color.

Uniformity:

LEDs have a more uniform horizontal light distribution than HPS sources, as shown in figure 56, where the photo of HPS lighting has bright streaks across the road, and the photo of LED lighting does not. Even though VI on the pedestrians was matched between HPS and LED lighting, less negative contrast was needed in LED lighting for pedestrian detection. That is because HPS lighting created bright lines across the roadway that increased background luminance values and therefore increasing the absolute contrast.

Figure 56. Photo. Uniformity of HPS and LED lighting.

Target Detection Distance

Luminance:

At the moment participants detected targets under HPS lighting, the average target luminance was 0.916 cd/m² (0.267 fL), greater than that in LED lighting, 0.683 cd/m² (0.199 fL). There appears to be a threshold at a luminance of approximately 0.2 cd/m² (0.06 fL) where targets were detected, shown in figure 57.

1 cd/m² = 0.3 fL
1 m = 3.3 ft

Figure 57. Scatter Plot. Scoping experiment—target-detection distance and luminance.

Contrast:

The average contrast when targets were detected was also different between the lighting types, with the contrast in HPS lighting at 0.788 and LED lighting at 0.734. Unlike the pedestrians, the targets were in positive contrast at detection. Similar to the pedestrians, a higher absolute contrast was needed in HPS than LED lighting to detect the targets.

Discussion

The study considered many factors, including overhead-lighting type (2,100-K HPS and 6,000-K LED) and intensity, and headlamp color and intensity.

A project objective was to measure the combined effect of vehicle headlamps and overhead lighting on visibility. Results of this experiment indicated that headlamp color and intensity did not significantly affect detection and color-recognition distances. While that result may be expected in situations where the overhead lighting is the predominant lighting source, when there was no overhead lighting present and the driver relied solely on headlamp light, there was still no statistically significant difference in visibility between the white/blue or white/yellow headlamps or between high- and low-powered headlamps. That result held true for detection and color-recognition for both targets and pedestrians. Interestingly, results were closest to approaching significance for targets positioned low to the ground in the dark section of the road. A possible explanation for this could be that the beam pattern of the headlamps is critical for highlighting low targets on the roadway.

The project also aimed to measure the impact of spectra of overhead-lighting systems on visibility, another project objective. The results regarding the effect of overhead-lighting type, (HPS with a correlated color temperature of 2,100 K and LED with a correlated color temperature of 6,000 K) and overhead-lighting intensity on pedestrian detection were mixed. For the constant-contrast pedestrian, there was no significant difference between the effects of overhead-lighting type and effects of level of the overhead lighting on detection distance. However, results combining all pedestrian locations indicated a very strong impact of overhead-lighting type and level on pedestrian-detection distance. This is an interesting result because it highlights the importance of contrast for detection. A few aspects of the experiment could have produced this result. First, constant-contrast pedestrians only wore gray clothing, so interactions between the spectral distribution of the overhead lighting and clothing color contrast would not be present. Second, this task was likely foveal, so mesopic effects would be minimal. The constant-contrast result shows that contrast is by far the more dominant impact on detection compared with other factors such as overhead-lighting SPD and intensity. It also demonstrates that the spectral effects are by far more evident in the periphery, as shown by the off-axis results and by the fact that changing the light source spectrum did not improve performance when all other factors were controlled.

Despite informing participants that pedestrians could be located both along the roadway and off to the side of the roadway, there were very few detections of off-axis pedestrians. The low detection rate for the off-axis pedestrians did not produce enough data to draw firm conclusions. However, for off-axis pedestrians, high-level LED overhead lighting resulted in a greater detection distances than high-level HPS overhead lighting; the same was not true for low-level lighting. The results were not statistically significant but are worth considering because they might indicate an effect of the overhead lighting’s uniformity. One possible reason for the greater detection distances in high-level LED lighting could be the distribution of the light source and how it illuminates the environment outside the roadway. The off-axis pedestrians stood on a grassy and rocky area, much less uniform than a typical roadway surface. The LED overhead lighting was more uniformly distributed than the HPS lighting. Thus, the more-uniform LED overhead lighting combined with the less-uniform background for off-axis pedestrians could have allowed participants to detect those pedestrians from farther away.

Conclusions

A research objective of this experiment was to evaluate the effect of the spectral distribution of overhead-lighting sources on drivers’ ability to detect pedestrians and targets and recognize colors in the environment. Results found that overhead-lighting type and level significantly affected detection and color-recognition distances for pedestrians, with brighter lighting corresponding to longer detection and color-recognition distances. The effect of a change in overhead-lighting level for a particular type of lighting was significant. For example, lighting level significantly affected detection and color-recognition distances for LED lighting. For HPS lighting, lower levels of lighting also resulted in shorter detection and color-recognition distances, but the effect was not significant. The different results for LED and HPS lighting indicate a spectral effect occurred with the LED lighting that did not occur with HPS lighting.

Another research objective for this experiment was to evaluate the impact of spectral distribution of overhead lighting and headlamp lighting on detection of pedestrians and targets located peripherally. Results indicate that, for off-axis pedestrians, LED overhead lighting had greater (though not statistically significantly greater) color-recognition distances than HPS lighting, possibly indicating a spectral effect of overhead-lighting type.

Evaluating the effect of the spectral distribution of vehicle headlamp color on drivers’ ability to detect pedestrians and targets and recognize colors in the environment was an additional research objective. Results found that headlamp color and intensity did not significantly affect detection and color-recognition distances when overhead lighting was used. Because overhead lighting had a greater effect on visibility than headlamps, overhead lighting was the focus of subsequent experiments.

Other noteworthy results include that age significantly affected detection distances when there was overhead lighting but not without overhead lighting. Age-related response times might have been a factor, but responses were not corrected for this factor. Also, age-related response times would not explain why age significantly affected detection distances in overhead lighting only. Age effects were not a main research objective, and these results were not further analyzed in this experiment.

Pedestrian clothing color and target color both significantly affected detection and color-recognition distances, whether or not overhead lighting was used. The best and worst colors for detection and color recognition varied between pedestrians and targets, and between whether or not there was overhead lighting, as listed in table 14. However, clothing and target color were not the focus of this project on light-source spectra.

Table 14. Scoping experiment best and worst colors for target and pedestrian detection and color recognition, with and without overhead lighting.

The uniformity of the overhead lighting appeared to affect the results. At the time of detection, pedestrians contrasted more against the background under HPS lighting than under LED lighting, likely because HPS lighting was less uniform and created bright lines across the roadway. Future experiments included HPS and LED lighting to further investigate the effect of the overhead lighting’s uniformity on object detection.

Design of Further Experiments

The results of the scoping experiment were used to inform the design of the subsequent phase of the project. This project’s subsequent investigation of the spectral impact of the overhead-lighting sources included the following efforts:

• Evaluate the full range of the spectral impact of the light source on the visibility of both on-axis and off-axis targets because the scoping experiment showed that a wider range of off-axis detection objects was required to more fully describe the impact of overhead lighting’s spectral effect.
• Test an MPI peripheral highlighting system’s effectiveness in increasing off-axis pedestrian visibility, possibly increasing detection rates from the very low 23-percent rate found in the scoping experiment.
• Evaluate further the interaction of overhead lighting and headlamps on visibility because results from the scoping experiment regarding this interaction were not statistically significant.
• Evaluate the effectiveness of the CIE-developed mesopic model for describing nighttime driving visibility conducted both in a static and a dynamic environment because the scoping experiment was not specifically designed to address this project objective.

Limitations

Participant fatigue and learning effects might have affected the results. There were too few off-axis detections to draw strong conclusions regarding off-axis detection.