Seeing Pedestrians At Night
The visibility of pedestrians at night is a complex phenomenon. Below, I outline some of the major issues. See Green et al (2008) for more in depth explanation.Pedestrians Usually Appear In Peripheral Vision
A Driver can only avoid an accident if he sees the pedestrian relatively far ahead. The driver must detect the pedestrian, recognize the impending collision, and then react by pressing the brakes. Suppose a driver is traveling at 60 mph (88 feet per second) and suddenly sees a pedestrian. "Normal" perception-reaction time1 for a lane incursion by a pedestrian (Green, 2000) is about 1.5 seconds. During this time, the car will have moved forward 132 feet (1.5 x 88). Cars don’t stop instantaneously, however, so the vehicle continues forward after brake depression until friction halts all motion. This would require another 150 feet and take about 3.4 seconds. As a result, the driver must see the pedestrian 280 feet and 4.9 seconds in advance. At city driving speeds of 35 mph, the stopping distance is still 138 feet and the stopping time is 3.6 seconds. These calculations also assume good conditions and ignore brake lag, the time from touching the pedal to full depression and lock up, which is likely .25-.50 sec. The requirements increase when the road is wet, the vehicle is heavier (a fully loaded tractor trailer can take 400 feet or more), the terrain slopes downhill or the driver’s reaction time is slowed by age, alcohol, fatigue or distraction. (Think cell phone.)
The figure below shows what occurs when a driver approaches a pedestrian. First, suppose the pedestrian is walking at a steady pace toward the roadway and across the car’s path.
The driver is usually looking directly ahead. The pedestrian must be seen in peripheral vision because there is an angle between the driver’s direction of gaze and the pedestrian’s location. The pedestrian is hard to detect because people are much poorer at seeing objects off the direct line-of-sight. The eye has lower peripheral acuity and visual attention is usually concentrated where the viewer fixates – directly ahead. Attention becomes more focused ahead when visibility conditions are poor or traffic is heavy.
Next, the driver travels forward and the pedestrian walks leftward. If the two are on a collision course (figure middle), the angle remains constant. The pedestrian never leaves peripheral vision and may go unseen. If the pedestrian speeds up (say, to beat a crossing signal) or the driver slows, the angle shrinks and then s/he will move toward the center of vision. The driver will then see the pedestrian suddenly appear out of nowhere. In some cases, this may be literally true if the pedestrian had been hidden behind a view-blocking windshield post.
The driver will most likely detect the pedestrian only if there is an eye movement to the right. This may occur by chance or if the pedestrian is highly conspicuous for some reason, but drivers are usually looking for pedestrians only at likely locations (Green, 2002). Drivers won’t expect to see pedestrians in rural roads or other areas where they have learned through experience that pedestrians are unlikely to appear. For example, three quarters of the pedestrians killed in road accidents are hit outside of crosswalks, probably because drivers do not expect them there.
In another common scenario, the driver approaches with the pedestrian remaining stationary on the sidewalk (right figure). As the distance between the two shrinks, the angle grows. The pedestrian moves farther into peripheral vision and becomes increasingly difficult to detect. Once the pedestrian starts walking or running, the situation turns into the middle case.
This scenario reveals an interesting anomaly; the best opportunity for the driver to see the pedestrian often occurs when the car is still far away. As the car approaches, the pedestrian recedes farther into the driver’s peripheral vision. At larger distances, the pedestrian is closer to the driver’s line of sight and is more detectable.
Unfortunately, there is a catch-22 because at greater distances the pedestrian’s image is smaller, which decreases visibility. If the driver is 4 seconds away from the pedestrian, the sight distance is about 350 feet at 60 mph and about 200 feet at 35 mph. These are great distances to see a pedestrian, possibly against a cluttered background.
In sum, the geometry of the situation, and the innate human visual loss for objects located in the periphery and for small objects, conspire to make pedestrian detection a difficult task. The situation is even worse at night because headlamps do not allow drivers to see very far ahead.Pedestrian Visibility Dims At Night
When scaled by the number of miles driven, pedestrian fatality rate is three times higher at night. Part of the reason is greater chance of driver drinking and fatigue, but the critical factor is lower visibility due to reduced ambient illumination. People have contrast sensitivity in dim light. Moreover, the eye exhibits “night myopia” focusing too near and causing distant objects to blur.
The obvious solution, increased highway lighting, is impractical due to high energy costs. Moreover, there would also be a great outcry against the resulting “light pollution.” Since road lighting cannot be made sufficiently high for daytime safety levels, drivers typically rely on headlamp illumination to detect pedestrians. However, normal low-beams make pedestrians visible only at relatively short distances. The key concept is “assured-clear-distance,” which refers to the distance ahead that a driver can see a pedestrian on the road. Most drivers are taught to drive slowly enough that they could stop for a pedestrian who just falls within their assured-clear-distance, otherwise they would be “overdriving” their headlamps. Some US states have even made this a matter of law; anyone who overdrives his/her headlamps and has an accident is automatically guilty. However, automobile headlamps provide such a short assured-clear-distance that even drivers who obey the speed limit are often overdriving their headlamps.
People fail to slow sufficiently at night because they are unaware how poor their vision has become. Humans have two distinctly different types of vision, focal and ambient. They differ in the visual tasks that they perform, the parts of the visual field they examine and their pathways through the brain. Roughly speaking, focal vision tells us “what is there” while ambient vision tells us “where we are”
Focal vision is used for detecting and recognizing objects, such as pedestrians. It is centered along the line of sight, so when we want to recognize an object, we turn our eyes to look directly at it. Focal vision declines rapidly in dim light. In fact, many people with impaired vision who are unable to receive driver’s license still have vision superior to that of a normal person at night.
Ambient vision is used for determining location in space and orientation in the environment and to perform tasks such as steering a car. It operates out in the visual periphery and needs only detect faint large shapes. Most significantly, ambient vision is not greatly impaired when light level declines. Drivers can steer the car just as well at night as during the day and feel little need to slow. They do not realize that their ability to see pedestrians has been greatly reduced.
Many studies have investigated the exact distance at which normal headlamps permit pedestrian detection. There is no single estimate because the distance is a function of many factors, including pedestrian clothing and location, and driver age and expectation.
In one study, the average driver saw dark-clothed pedestrians standing “a foot or two” to the right of the car at a range of 150 feet while 90% of the drivers fell in a range from 50-250 feet. When pedestrians stood to the left, the visibility distances halved. Using some reasonable assumptions, the authors concluded that a driver traveling 55 mph would fail to see a pedestrian on the right in time to avoid collision 45% of the time. If the pedestrian is standing to the left, the number grows to 95% - the driver will almost always hit the pedestrian.
Switching from dark clothing to a white vest markedly improved visibility. Drivers then detected pedestrians on the right at about 300 feet and pedestrians on the left at about 200 feet. Theoretically, the vest reduces the number of accidents to 3% right and 9% left, or an improvement by a factor of 10.
The benefit of lighter clothing is to be expected since detection depends on the amount of light reflected from the pedestrian back to the driver’s eye. This, in turn, is the amount of light falling on the pedestrian multiplied by reflectivity of the clothing. Pedestrians could be made more visible either by being more brightly illuminated by a closer headlamp or by wearing lighter, more reflective clothing.
Driver age is another important variable. A group of older drivers detected the pedestrians at only 60% of the distance of a younger group. Older drivers traveling 55 mph would then hit almost all dark-clad pedestrians. Even at 35 mph, there would be insufficient time to stop for half of the pedestrians standing on the right.
Although these predicted accident rates may seem alarming, the reality is even worse. Most studies use drivers who were told to look for a pedestrian ahead. They expected to see the pedestrian and knew the approximate location. Real drivers have no such knowledge and certainty. When drivers are tested without being told that a pedestrian would be ahead, visibility distances decreased by up to 50%. As a rule of thumb, unalerted drivers will seldom see a pedestrian at distances much greater than 100 feet. Recall from above that drivers traveling at city speeds of 35 mph (about 60 kph) take about 138 feet to stop. They are overdriving their headlamps by a large margin.Headlamps Aim In The Wrong Direction
Given the dangerously short assured-clear-distance that headlamps provide, it might be expected that manufacturers would increase headlamp power. This is not possible, however, because headlamps must operate under an important constraint – while making pedestrians more visible, they must not put blinding glare in the eyes of oncoming motorists. Headlamp brightness must be limited, lowering the distance at which they could potentially illuminate pedestrians.
Further, the headlamps must aim away from the eyes of the oncoming drivers. The figure shows a typical beam pattern, the spread of light leaving the car from a properly aligned headlamp. The horizontal and vertical lines intersect at the center of the view ahead and the red x marks a circular area that is the highest intensity – the beam’s center. The roughly circular areas expanding outward indicate areas of equal illumination. The farther from the beam center, the lower the light level.
The diagram shows that the beam’s center aims down and to the right hand quadrant. An oncoming driver’s eyes would be located in the upper left-hand quadrant, so the beam will not hit the oncoming driver square in the eyes and cause blinding glare.
The down-right aim, however, greatly decreases pedestrian visibility. The beam hits the road close to the vehicle, rather than illuminating objects at a greater distance. Moreover, pedestrians approaching from the left will be far less visible than those approaching from the right. I have already noted that studies find visibility distance is much shorter for pedestrians located to the vehicle’s left.
The downward aim also causes headlamps to illuminate pedestrians in the worst possible location, near the ground. To see why, consider where the light from car headlamps hits the pedestrian and background (see Figure). Drivers detect pedestrians by their contrast, the difference in brightness between pedestrian and background. In an ideal world, headlamps would maximize contrast by illuminating the pedestrian but not the background. Headlamp illumination falls with distance, so the farther the background from the pedestrian, the less light it receives and reflects back and the more the ideal situation is approximated.
The figure shows that the beam aim creates poor lighting conditions. Point A, the pedestrian’s leg, is only slightly in front of point B, the road just behind the leg. The down-aiming car headlamp then illuminates the background almost as well as the pedestrian’s lower body. Visibility depends on the difference in reflectance between the pedestrian’s pants and the roadway, so they receive similar intensity light.Now look at point C high up on the pedestrian’s body. The background is far away, so it does not reflect significant headlamp illumination. The contrast between object and background will then approach the ideal situation. Unfortunately, the downward pointing headlamps throw the most light where it does the least good - where the distance between pedestrian and background is small - and illuminates the least where it does the most good - where the distance between pedestrian and background is large.
This partly explains why children are difficult to see on the road. They are closer to the ground, so they must be detected in a situation where the headlamps illuminate pedestrian and roadway similarly. Of course, the problem is compounded by their small size and by their walking gait, which drivers do not readily recognize as belonging to a person.
Lastly, many factors can reduce or alter headlamp illumination. Headlamps are often misaligned. Even the weight distribution in the car can change headlamp aim. For example, heavy objects in the trunk lower the car’s rear and cause the headlamps to aim upward. This may help the driver see farther, but the resulting glare is sure to reduce pedestrian visibility for the oncoming motorists who are blinded. Conversely, heavy objects in the front seat cause headlamps to aim farther downward. Research also shows that most people are driving cars where dirt is significantly reducing their headlamps' effective output. The problem is especially great in rain, when dirty water splashes up from the road coats the headlamps.So far, it sounds as if the more light that headlamps throw on the pedestrian, the greater the visibility. Further, light colored clothing should increase visibility since more light is reflected back to the driver’s eye. This view, however, is an oversimplification because increased light levels can actually decrease pedestrian visibility.
The reason is that people don’t see light; they see contrast. A driver may see a pedestrian in either positive or negative contrast. Negative contrast occurs when a dark object lies on a bright background (left) while Positive contrast occurs when a bright object lies on a dark background (right). In the driving literature, negative contrast is often called “silhouette” and positive contrast is termed “reverse silhouette.”
In real road situations, positive and negative contrast can compete, rendering the pedestrian invisible. Imagine a car on a dark rural road at night. The driver will likely see a pedestrian in positive contrast because the light from the headlamps reflects off clothing back to the eye and because the background is dark - there is little or no background lighting. In this case, pedestrian clothing is an important factor in visibility. If wearing light clothes, then most of the headlamp illumination reflects back to the driver’s eye. Visibility will be far lower for pedestrians wearing dark clothes that reflect less light. The studies discussed above all used dark roads where positive contrast would be maximized.
Negative contrast occurs when the prevailing light comes from behind. If there is backlighting from streetlamps, stores, sky glow, headlamps behind, moon, etc, then a pedestrian may appear as a silhouette against the lighted backdrop. In this case, the pedestrian’s outline is black regardless of clothing. This might seem fortunate because viewers have slightly greater contrast sensitivity to negative contrast. However, it is probably more difficult to notice an object with negative contrast. When people first view a scene, they perform "figure-ground" segregation, the task of breaking the scene into foreground objects that deserve against a background which does not. Viewers tend to judge bright parts of the scene as foreground, which makes dark object unnoticed background. A dark pedestrian is especially a hole in the scene. Moreover, viewers are better at noticing and identifying objects if their internal 3D structure is visible. The silhouette seen in negative contrast has no visible internal structure.
Complications arise because positive and negative contrast can cancel one another. For example, light colored clothing decreases negative contrast by better reflecting any ambient front light back to the driver’s eye. With negative contrast, dark clothing creates better visibility. Finally, a pedestrian wearing different color top and bottom clothing can appear as two smaller objects, one of positive contrast and the other of negative. This further impairs pedestrian visibility because 1) each piece is small and therefore less detectable and 2) the object is more difficult to identify as being a pedestrian.
The dominant contrast polarity often depends on viewing distance. Since car low beam headlamps provide usable light for a distance of only 100 feet, only shorter distances promote positive contrast. At longer distances, negative contrast is favored if there is much backlighting. The pedestrian then switches from negative to positive contrast as the driver approaches. At some point during the transition, however, the pedestrian must go through a period of zero contrast and be literally invisible.
The conflict between positive and negative polarity also explains why twilight is an especially dangerous time for pedestrians. The overall illumination decreases quickly, so human contrast perception degrades rapidly. Yet, there is still some ambient illumination to provide both front and back lighting and hence both positive and negative contrast. As a result, the driver has reduced contrast vision while the positive and negative contrasts are canceling one another.
Lastly, negative contrast can be good for providing visibility, but it is poor for conspicuity and for recognition. In positive contrast, the viewer can distinguish the pants, shirt, face, etc. and can more readily recognize the object as a person. Negative contrast provides just an outline that has few cues as to the object's meaning, which is important for capturing attention. This recognition problem will be especially bad for a pedestrian walking in the same or opposite direction to the car. In this case, the outline is simply an oblong with few form or motion cues that would mark it as a person.Pedestrians Help Make Themselves Invisible
Pedestrians are themselves partly to blame for their invisibility and high accident rate. I’ve already noted that pedestrians frequently confound driver expectation by crossing roadways at arbitrary locations. Further, people also generally prefer dark clothing. Safety authorities often suggest that pedestrians would become more conspicuous if they would wear reflective material that sends more light back to the driver’s eye. Research typically confirms that pedestrians are visible at greater distances when they wear a reflective tag or vest. However, there are some drawbacks to reflective material. One is that reflective material sends light primarily in one direction. If the headlamps hit the material at the wrong angle, the reflected light goes in the wrong direction and does not hit the driver’s eye, and the reflector will appear dark. Further, if the reflective material covers a small part of the body, then the driver may detect its light but not recognize it as being a person.
Reflective material may also cause pedestrians to be overconfident. Pedestrians already greatly overestimate their visibility at night. They see a highly conspicuous car coming toward them and imagine that they must be just as visible to the driver as the car is to them. Studies show that pedestrians typically overrate their visibility distance by a factor of two.
The use of reflective material is likely to amplify pedestrian overconfidence. Introduction of new safety devices often makes people feel more secure, so they engage in riskier behavior (Green, 2001a,b). A pedestrian wearing reflective material may be more likely to assume high visibility and take more risks.
Lastly, many, if not most, pedestrians killed in automobile accidents had been drinking or taking drugs. Even moderate doses of alcohol impair perception, attention and motor skill (the ability to move quickly). Accident statistics reflect these effects by showing that alcohol ingestion greatly increases both the likelihood that a pedestrian will be involved in an automobile accident and the severity of the resulting injury. One study found that ¾ of all pedestrians killed by automobiles had alcohol in their blood streams while another concluded that alcohol increases the odds of a pedestrian being involved in a fatal accident by a factor of 5. In contrast, drinking and driving increases accident rates by a factor of “only” 4. Drunken walking may be even more dangerous than drunken driving!
The reasons for pedestrian invisibility are complex, but most lie in the normal operation of human vision and attention. For a driver to avert accident, the pedestrian must be seen far ahead, a time when he is likely to appear in peripheral vision At night, the problems are exacerbated by the lack of ambient illumination, the limited power and aim of headlights, the conflict between positive and negative contrast and by night myopia and glare. Lastly, pedestrians contribute significantly to their own invisibility by choice of clothing, and risk-taking due to overconfidence in their conspicuity and to drinking or taking drugs.
Footnotes1At most, this is "normal" perception-reaction time that is expected only during good visibility in daylight condition, in good weather, with normal vision, etc., etc. At night, reaction time is likely longer, possible as long as infinity. Perception-reaction time is a very complex phenomenon, which is affected by many variables. In sum, there is no such thing as a global "normal" value. See Green (2009) for more details.
Green, M. (2000). ’How long does it take to stop?’ Analysis of brake reaction times,” Transportation Human Factors, 2, 195-216.
Green, M. (2001a). The psychology of warnings. Occupational Health and Safety Canada, 30-38, October/November, 30-38.
Green, M. (2001b). Do Mobile phones pose an unacceptable risk? A Complete look at the adequacy of the evidence? Risk Management, November, 40-48.
Green, M. (2002). Inattentional Blindness, Occupational Health & Safety Canada, Jan/Feb, 23-29.
Green, M. (2009). Perception-reaction time: Is Olson (& Sivak) all you need to know? Collision, 4, 88-95.
|| Home | Experience | Services | Contact Us | Seminars/CLE | Attorney's Guide | Resources ||
Copyright © 2009 Marc Green, Phd
Home Page: http://www.visualexpert.com