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-04-132
Date: December 2005
Enhanced Night Visibility, Volume I: Executive Summary
PDF Version (233 KB)
PDF files can be viewed with the Acrobat® Reader®
CHAPTER 3—PHASES II AND III VISUAL PERFORMANCE STUDIES
To measure drivers' visual performance using different types of VESs, nighttime experiments required volunteer participants to drive various types of vehicles outfitted with a variety of headlamps and combinations of headlamps with supplemental UV–A or IR technology. The various VESs tested were obtained either through coordination with potential manufacturers and suppliers or by purchasing and combining off-the-shelf components. The participants were asked to report when they could detect, and then recognize, different objects placed on or near the roadway. Test vehicles were instrumented to record the distance to the object at the moment of detection and recognition. The testing was conducted on the Virginia Smart Road, 3.2 km (2 mi) of two lanes of roadway (closed to public traffic) that includes weather-making capability. Separate studies tested object detection and recognition in clear and adverse weather conditions—rain, snow, and fog.
The visual performance studies included three independent variables: VES, age, and object. The VES variable included different types of headlamps as well as headlamps combined with supplemental UV–A or IR systems. The age variable grouped participants into three age groups. The object variable included various roadway objects and pedestrians that participants were required to detect and recognize.
The VESs used for the visual performance studies included several different technologies. The studies reported in ENV Volume III (clear weather), ENV Volume IV (rain), ENV Volume V (snow), and ENV Volume VI (fog) included the following VESs:
The studies reported in ENV Volume XIII (clear weather) and ENV Volume XIV (rain) included these VESs:
The Phase II research used the following three UV–A configurations: two hybrid UV–A lamps (hybrid UV–A), so called because of their significant visible light component in conjunction with the UV–A component; three UV–A lamps (three UV–A) that had a minimal visible light component; and five of these UV–A lamps (five UV–A). It is important to recognize that the five UV–A headlamp configuration was included to provide a proof-of-concept, evaluating the maximum potential benefits of a UV–A supplemental headlamp system. As described in detail in ENV Volume XVII, the configuration used five large, high-wattage lamps designed for use on snowplows in Norway. Thus, barring significant advances in technology (such as UV–A light-emitting diodes), providing this much UV–A light is not practical for installation on automobiles at this time because of the UV–A headlamps' cost, power consumption, and size.
Each of the UV–A configurations was paired with halogen headlamps and, separately, with HID headlamps.
The research in Phase III evaluated two NIR systems, which used IR emitters in combination with a camera sensitive to the near IR spectrum. Both Phases II and III evaluated an IR–TIS system, which used a camera sensitive to thermal contrast between objects and surroundings. A display located just above the instrument panel presented images from these systems. Because NIR and IR–TIS are supplemental visibility systems not designed to be used without visible light in vehicular applications, each system was paired with halogen headlamps provided by the IR system suppliers; these halogen headlamps were different from the halogen headlamps paired with the UV–A headlamps.
Headlamps were aimed before starting the experiment's session each night. At the beginning of the project, a headlamp aiming device was not available to the contractor, so an aiming protocol was developed with the help of experts in the field. During the photometric characterization of the headlamps, it was discovered that the position of the maximum intensity location of the HLB, HOH, and HHB configurations was aimed higher and more toward the left than typical. This aiming deviation likely increased detection and recognition distances for the HLB and HOH configurations and likely decreased them for the HHB configuration. Details about the aiming procedure and the maximum intensity location are discussed in ENV Volume XVII, Characterization of Experimental Vision Enhancement Systems.
All of the studies except for the snow study used three age groups: younger participants (18 to 25 years), middle-aged participants (40 to 50 years), and older participants (65 years or older). The older group was excluded from the snow condition because the participants were required to get in and out of the experimental vehicles multiple times throughout the night on a potentially icy road surface. The risk for a slip and fall, although unlikely, was deemed too great to allow older drivers to participate.
The objects used for these studies were selected to represent a variety of potential roadway obstacles and pedestrian scenarios. Table 1 shows all the objects included in this project as well as the weather conditions and phase of the project in which they were used. All the objects were static with the exception of the parallel pedestrians, perpendicular pedestrians, and cyclists. Parallel pedestrians continuously walked back and forth along the shoulder next to the road's right edgeline. Perpendicular pedestrians continuously walked from the right edgeline of the road to the centerline and back. Cyclists continuously rode from one side of the road to the other. All other objects were statically positioned on the side of the road near the edgeline or, in the case of the far off axis pedestrians, 9.4 m (31 ft) to the left or right of the centerline. "Bloom" pedestrians stood beside a car parked in the oncoming lane with its headlamps on. In this configuration, the glare from the oncoming vehicle had the potential to overload the NIR camera. This caused the image in the system display to be washed out with a bloom of light and caused the pedestrian to be nonvisible in the display.
This project used real people for the pedestrians to allow for movement and to give the IR–TIS system a realistic heat differential. The dog was an internally heated, stuffed model of a Scottish terrier.
The primary performance variables were the distance at which participants detected an object and the distance at which they recognized an object. Detection was explained to the participants as follows: "Detection is when you can just tell that something is on the road in front of you. You cannot tell what the object is, but you know something is there." Recognition was explained as follows: "Recognition is when you not only know something is there, but you also know what it is." Later, participants were also asked to indicate their degree of agreement with a series of statements that addressed their perceptions of improved vision, safety, and comfort after using each VES. Participants rated their agreement to these questions using a seven-point Likert-type scale with anchor points at "1," indicating "Strongly Agree," and "7," indicating "Strongly Disagree."
The Phase II results indicate that VESs with supplemental UV–A generally did not provide sufficient improvement over the tested HID and HLB headlamps to justify additional research in this area. As expected, the UV–A produced longer detection distances of the pedestrian dressed in white for the clear, rain, fog, and snow conditions than did the HID or HLB headlamps tested alone. The UV–A also provided longer detection distances of all the objects on average; however, even five UV–A, the most powerful UV–A configuration, provided improvements ranging from only 7 m (23 ft) in adverse weather to 16 m (52 ft) in clear conditions. Given these small benefits combined with the current impracticality of producing this much UV–A with a vehicle, the UV–A conditions were excluded from the Phase III research.
The Phase II experiments showed that for the pedestrian dressed in white in adverse weather, IR–TIS showed a 12 m (39-ft) improvement over the baseline HLB headlamps in fog, but in heavy rain it showed a 6 m (20-ft) decrement; the system was not used in the snow condition because snow buildup would have blocked the camera. In the Phase II clear condition research comparing the headlamps supplemented with IR–TIS to the headlamps alone (ENV Volume III), the supplemental IR–TIS showed an approximately 55 m (181-ft) greater detection distance for the pedestrian dressed in white and a more than 100-m (328-ft) greater detection distance for the pedestrian dressed in black. This latter finding was important because pedestrians often wear low-contrast, nonreflective clothing.(5) The overall Phase II results show that IR technology has the potential to reduce pedestrian crashes by increasing detection and recognition distances; therefore, the emphasis for this project was shifted from testing UV–A to testing IR technology and other headlamps more thoroughly in Phase III. As the Phase II research came to an end, automobile manufacturers became more interested in near IR, which has the potential to greatly increase detection distance in inclement weather. For this reason, the Phase III testing included the IR–TIS system as well as two prototype NIR systems.
The overall Phase III results indicated that both the IR–TIS and one of the NIR systems could outperform headlamps alone in pedestrian detection (ENV Volume XIII). For most of the pedestrian scenarios, the IR–TIS implementation provided a 20- to 30 m (66-ft to 98-ft) detection advantage over the best NIR implementation. The second NIR system did not perform as well as the other two IR systems or even some of the headlamps, illustrating that implementation is the key to a successful enhanced night vision system. Both NIR and IR–TIS improved detection distance of pedestrians, compared to visible headlamp systems, when a glare source was present.
In general, in rainy driving conditions (ENV Volume XIV), both NIR systems had longer detection distances than the HLB and the IR–TIS system for nearly all pedestrian detection scenarios. This is a particularly interesting finding because even the NIR system that did not perform well in the clear condition outperformed the other systems in adverse weather. These objective findings do not appear to be differentiated by age and are corroborated by the subjective responses of the drivers in this study.
In the Phase II clear condition study (ENV Volume III), older drivers had shorter detection distances on average than the younger and middle-aged drivers, especially with low-contrast objects, but the differences were smaller with the IR–TIS. Supplemental IR showed this benefit for older participants in Phase III also (ENV Volume XIII). A more detailed analysis of this Phase III data showed that the older participants using the IR systems performed similarly to the younger participants using the best of the three headlamp systems, indicating that these IR systems could be used to reduce an age-related decrement in object detection.
In the Phase II clear and rain conditions (ENV Volumes III and IV), clothing contrast rather than object motion appears to have been responsible for differences in detection distances observed between the different types of pedestrians and cyclists. Not surprisingly, pedestrians dressed in white were detected farther away than pedestrians dressed in black regardless of the VES used.
In the Phase III clear condition study (ENV Volume XIII), on average the VESs demonstrated longer detection distances of pedestrians dressed in blue clothing than of pedestrians dressed in black clothing by 60 percent. Although this result was not surprising for most of the VESs, the 83 m (273-ft) greater detection distance for blue clothing when using an IR–TIS system was greater than expected. IR–TIS imaging is based on thermal differences between the object and the background rather than differences in the visible spectrum, so theoretically there should have been no difference in pedestrian detection because of clothing color. The observed difference could have been the result of the thicker blue cloth retaining more of the pedestrians' body heat than the thinner black cloth, or perhaps some participants waited for visual confirmation (through the windshield) before declaring detection of a pedestrian.
In Phase II, the drivers' subjective evaluations suggest that they thought HID helped them detect and recognize the different objects from farther away than did the other VESs. This finding conflicts with the objective data, which show shorter detection and recognition distances with HID. This conflict indicates that collecting subjective data alone for this type of research is not sufficient.