ArticlePDF Available

Why fog increases the perceived speed

Authors:

Abstract and Figures

In the first experiment we investigated the effect of reduced visibility on the produced speed in a driving simulation. Participants were required to drive at a target speed of 90 km/h in different visibility conditions. We found that when realistic fog was simulated, the driving speed was reduced accordingly to the fog density. When a uniform reduction of the image contrast was implemented, no effects were observed on the produced speed. We speculated that fog reduces selectively the visibility of the distant region of the scene and leaves visible only the proximal area that contains high angular velocities. We hypothesized that the perceived speed is then biased by the available raw velocity signals from the visual field. In the second experiment we addressed the question whether the observed behavioral effect has indeed a perceptual origin. In a psychophysical task we asked the participants to estimate the speed of moving scenes when the sight was limited either to the periphery (high angular velocities) o r to the center (low angular velocities) of the field of view. According to our hypothesis, we found that when the central region was occluded, the speed at the periphery was perceived as being higher, and conversely, when the peripheral region was missing the speed at the center was perceived as being lower. We conclude that the speed reduction while driving in fog is due to a non-optimal perceptual compensation for the hidden central region with low angular velocities, which causes an overestimation of the driving speed.
Content may be subject to copyright.
WHY FOG INCREASES THE PERCEIVED
SPEED
Paolo Pretto1, Manuel Vidal2, Astros Chatziastros1
1Max Planck Institute for Biological Cybernetics
Spemansstr. 38, 72076 Tübingen
Germany
2LPPA, Collège de France CNRS, Paris
DSC 2008 Europe Monaco 31 January - 1 February 2008
Abstract
In the first experiment we investigated the effect of reduced visibility on the produced speed
in a driving simulation. Participants were required to drive at a target speed of 90 km/h in
different visibility conditions. We found that when realistic fog was simulated, the driving
speed was reduced accordingly to the fog density. When a uniform reduction of the image
contrast was implemented, no effects were observed on the produced speed. We speculated
that fog reduces selectively the visibility of the distant region of the scene and leaves visible
only the proximal area that contains high angular velocities. We hypothesized that the
perceived speed is then biased by the available raw velocity signals from the visual field. In
the second experiment we addressed the question whether the observed behavioral effect has
indeed a perceptual origin. In a psychophysical task we asked the participants to estimate the
speed of moving scenes when the sight was limited either to the periphery (high angular
velocities) or to the center (low angular velocities) of the field of view. According to our
hypothesis, we found that when the central region was occluded, the speed at the periphery
was perceived as being higher, and conversely, when the peripheral region was missing the
speed at the center was perceived as being lower. We conclude that the speed reduction while
driving in fog is due to a non-optimal perceptual compensation for the hidden central region
with low angular velocities, which causes an overestimation of the driving speed.
Résumé
DSC 2008 Europe Monaco 31 January - 1 February 2008
1. Introduction
Estimating the speed of one own vehicle and that of the other objects in the environment is an
activity that is done continuously while driving. Many studies have been carried out in order
to understand how drivers perceive the speed and what are the factors that influence the
driving behavior (Denton, 1980; Evans, 1991; Recarte & Nunes, 1996). (2003)
Psychophysical studies on two-dimensional motion perception have shown that observers can
misestimate speed when image texture (Blakemore & Snowden, 2000) or luminance
(Takeuchi & De, 2000) are reduced. An extensively studied perceptual factor that has been
shown to affect the perceived speed of moving objects is the contrast of the scene (Thompson,
1982; Stone & Thompson, 1992; Hawken, Gegenfurtner, & Tang, 1994). For instance, a
reduction of the contrast of two-dimensional moving patterns creates the visual illusion of a
decreased moving speed (Anstis, 2001). It has been argued that a similar phenomenon occurs
also in real life when driving in low visibility conditions (Green, 1983; Distler & Bülthoff,
1996; Anstis, 2003) or during the night (Gegenfurtner, Mayser, & Sharpe, 1999). However, it
has to be considered that in a driving scenario we experience self-motion in depth, rather than
mere translational motion of objects on a plane. In fact, every time we move through the
environment the images on the retina move accordingly. The retinal projection of forward
self-motion consists of an expanding optic flow where the angular velocity of the objects on
the scene depends on their position and speed relatively to the moving observer (Gibson J. J.,
1950). For instance, during driving, the closest regions in the environment results in higher
angular velocities, while lower angular velocities are usually in the centre of the visual field
and relate to distant regions. In a highly cited experiment, Snowden, Stimpson and Ruddle
(1998) have demonstrated that in a driving simulation with a uniform reduction of the contrast
drivers perceived a lower speed and tended to speed up in order to reach a given target speed.
However, with a realistic simulated fog instead of a uniform contrast reduction, we have
shown recently that the produced driving speed is decreased, as a consequence of a higher
perceived speed (Pretto & Chatziastros, 2006). Our contrast reduction affected differently the
distant and close regions of the environment and, therefore, of the visual field. The
explanation that we provided assumes that the fog reduces the visibility of the distant region
of the scene (slow angular velocities) so that only the information from the periphery of the
visual field (high angular velocities) is available. Therefore, the own speed estimate will be
biased towards higher values. This hypothesis is supported also by a behavioral study where it
has been shown that a restricted visual field decreases the self-rated perceived speed (Osaka,
1988).
In a first experiment we examined further the effect of both realistic fog and uniform contrast
reduction on the driving speed. In particular, we compared the effects of an increasingly
reduced visibility on the driving behavior in a realistic driving simulation.
In a second study we tested whether drivers estimate speed differently when sight is limited
either to the periphery or to the central portion of the field of view. With this perceptual
experiment we could address directly the question whether the difference in angular velocities
potentially produces a systematic bias in drivers’ estimation of their own speed.
2. Behavioral experiment
DSC 2008 Europe Monaco 31 January - 1 February 2008
In the first experiment we tested the speed production behavior of the drivers under clear and
reduced visibility conditions in a driving simulation with natural field-of-view (FOV).
2.1. Method
2.1.1. Apparatus and stimuli
The experimental setup consisted of a large (7m) semi-spherical screen, which depicts an
almost natural FOV (230° horizontal × 125° vertical), including a floor projection (figure 1).
The scene was displayed at 60Hz by four LCD-projectors with a resolution of 1400×1050
pixels each. Overlapping regions were blended by standard blending techniques. The
geometry correction of the projected scene was adjusted for an eye height of 0.8m at a
distance of 3m from the screen. A simplified vehicle mock-up equipped with steering wheel
and pedals served as driving interface.
Figure 1: Lateral (a) and top (b) view of a schematic representation of the experimental setup.
Dashed lines indicate the cones of projection of the four beamers. Glowing lines indicate the
extent of the surface where the scene was projected.
The virtual environment consisted of a model of a local dual-carriageway road with two lanes
in each direction. The road path had several curves with different radius (ranging from 250 to
7500 meters) and ascending and descending sections. A traffic divider on the left side
separated the two carriageways and road poles on the right side were placed every 50 meters.
Road markings on the textured road surface delineated the traffic lanes (figure 2).
DSC 2008 Europe Monaco 31 January - 1 February 2008
Figure 2: The experimental setup with the virtual environment displayed on the screen as it
appeared during the experiment.
In this highly realistic driving scenario we tested different visibility conditions by a
combination of two experimental factors: the type of contrast reduction (gradient or uniform)
and the amount of contrast reduction (no reduction, low, medium and high). In the conditions
with gradient of contrast reduction the visibility of the scene was reduced exponentially as a
function of the distance from the observer. This condition was implemented by adding an
exponential fog in the virtual scene, according to the fog model:
Cr = Cf (1 β) + Co β, (1)
where
β = e f d
and Cr is the resulting color, Co is the original color of the object displayed in that pixel, Cf is
the color of the fog, f is the density of the fog, and d is the linear distance of the object
displayed in that pixel.
The uniformly reduced contrast condition was implemented by inserting a transparent virtual
plane in front of the scene. The plane color was composited with the color of the background
image, according to a standard color transparency model (RGB-alpha):
Cr = Cp α + Co (1 α), (2)
where Cr is the resulting color, Cp is the color of transparent plane, Co is the original color of
the object displayed in that pixel, and α is the alpha value of the pixel in the transparent plane.
The color of both the fog and the plane was set to a medium gray level (RGB =
[128,128,128]), which represents the frequently experienced fog color due to light absorption
in the atmosphere. Figure 3 shows how the colors of the original scene were affected by
gradient and uniform contrast reduction.
Figure 3: Effects on the image colors in the gradient and uniform contrast reduction. In the foggy
conditions, a white pixel reduces its brightness, approaching the fog color (gray), as a function of
its distance from the observer. The density of the fog affects the steepness of the curve and the
distance at which the color of the scene become undistinguishable from the fog color (a). In the
DSC 2008 Europe Monaco 31 January - 1 February 2008
uniform contrast reduction, an increase in the opacity (alpha) of the transparent plane causes a
compression of the original pixel color toward the plane color (b).
We varied the amount of fog thickness in order to create four scenes where the visibility
ranged from clear to very low, according to the meteorological visibility range (MVR)
(Koschmieder, 1924), as shown in table 1. In the uniformly reduced contrast conditions, the
transparency of the plane was adjusted in order to obtain the same amount of contrast as
computed in the corresponding fog conditions. As suggested by Moulden, Kingdom and
Gatley (1990), we quantified the contrast of the scene as the normalized root mean square of
the luminance values of the pixels in the displayed environment (CRMS: the ratio of standard
deviation to mean). The colour of each pixel was converted into the corresponding brightness
(a gray-scale, achromatic value in the RGB colour system) and the scene luminance
distribution was computed based on the empirically determined function between luminance
and brightness. In table 1 the combination of values of the experimental conditions is shown,
together with the corresponding measured contrast.
Amount of contrast reduction
No
Low
Medium
High
Type of contrast
reduction
Gradient
(fog density)
0
0.10
0.20
0.30
Uniform
(plane opacity)
0
0.40
0.55
0.64
Measured contrast
CRMS
%
0.86
100
0.47
55
0.25
29
0.16
19
MVR
meters
30
15
10
Table 1: Experimental conditions and contrast levels. For each type and amount of contrast
reduction the corresponding values of the manipulated parameters (fog density and plane opacity)
are shown (note that opacity = 1 transparency). The contrast is reported as both dimensionless
number and percentage. The MVR indicates at which distance a white object appears with 5%
contrast.
In order to prevent drivers from decelerating because of unexpected curves, the white road-
edge lines were always unaffected by the fog and fully visible at any distance.
2.1.2. Participants
We tested 14 participants between the ages of 21 and 25. They were naïve as to the purposes
of the experiment, although they had been introduced to the topic of the research during a
preliminary meeting. All participants had normal or corrected-to-normal vision.
2.1.3. Design and procedure
In a within-subject experimental design we asked the participants to drive in all the 8
visibility conditions (2 type of contrast reduction × 4 amount of contrast reduction) in 5
repeated blocks, for an overall number of 40 trials. The conditions were randomly interleaved
within every block. The whole experiment lasted about one and a half hours.
A first training phase allowed the participant to familiarize with the task and the simulator.
During this phase the participants were required to drive for ten minutes in clear visibility
DSC 2008 Europe Monaco 31 January - 1 February 2008
conditions at the target speed of 90 km/h. Every time and as long as the driving speed and the
target speed were not matching, the current speed was being visualized on the screen. When
the correct speed was produced the numerical feedback disappeared. Participants were
instructed to learn how fast the scene looked like when driving at the right speed, in order to
reproduce the same speed in the following phase. In the experimental phase participants were
asked to drive on the right lane of the road, reach a target speed of 90 km/h, keep it for five
seconds, and terminate every trial by pressing a button. The average speed produced during
the last five seconds of each trial was considered as the produced speed for the following
statistical analyses.
2.2. Results
We found that the average produced speed for all the participants was about 96.6 km/h. At the
one-sample t-test this value resulted to be significantly higher than the target speed (t13 =
2.43, p < .05). In both type of contrast reductions a clear visibility condition was included. For
further analyses we rearranged the conditions in a new visibility factor with three levels (clear,
gradually reduced and uniformly reduced contrast). The results of this set of data are shown in
figure 4.
Figure 4: The produced speed in different visibility conditions. (a) The gradually reduced visibility
condition causes the lowest produced speed, whereas the other two conditions do not differ
significantly from each other. (b) None of the uniformly reduced visibility conditions differ
significantly from the clear visibility condition or from each other. Dashed lines indicate the two
original sets of data without distinction between clear and reduced visibility.
We conducted an analysis of variance for repeated measures (ANOVA) and found that the
visibility factor showed a significant main effect (F(2,26) = 6.24, p < 0.05). A post-hoc
analysis revealed that only the produced speed under gradually reduced visibility condition
was significantly lower compared to the clear condition. In the uniformly reduced visibility
condition the produced speed was not affected when compared to the clear visibility condition
(figure 4a). The interaction effect between type and amount of contrast reduction was not
significant (F(2,26) = 1.96, p = 0.16) (figure 4b).
*
DSC 2008 Europe Monaco 31 January - 1 February 2008
2.3. Discussion
A general speed overproduction is consistent with our previous study (Pretto & Chatziastros,
2006) and also with other studies that have reported the same phenomenon as compensation
for speed underestimation both in real and simulated driving conditions (Denton, 1980; Casey
& Lund, 1987; Kemeny & Panerai, 2003).
We did not observe an increase of the produced speed in the uniformly reduced contrast
conditions. The amount of contrast reduction used in our experiment is comparable to that of
the original study by Snowden et al. (1998). We then exclude that the effect of speed
overproduction was missing because of a weak contrast manipulation. It might be that the
extent of the field of view and the richness of the environment, in terms of available motion
cues, made the speed estimation process robust enough against the bias induced by the
contrast reduction. In particular, the extent of the lateral field of view could have allowed the
objects in the scene to stay visible for longer time. And this could have increased the
reliability of the perceived speed.
We observed, in accordance with previous findings (Dyre, Schaudt, & Lew, 2005; Pretto &
Chatziastros, 2006), that in the foggy scenario the participants produced a lower speed. This
result supports the explanation stating that fog masks distal portions of the scene, leaving only
the proximal parts with higher angular velocities visible. Furthermore, the trend of the
produced speed provides additional support to this explanation. In fact, an increment of the
fog density reduces the visible areas to the closer regions and we might then expect that the
speed rate is increased accordingly.
3. Psychophysical experiment
In the previous experiment we have shown that the behavior of participants changes when
driving in fog. From this result we could infer that the speed was perceived differently
according to the different visibility conditions. In this second experiment we intend to prove
that the measured behavioral effect is caused by a change at the perceptual level. Therefore
we use a psychophysical methodology, which does not include a speed production task but
rather introduces manipulations of the visual field, in order to examine directly how the speed
is perceived (Brandt, Dichgans, & Koenig, 1973).
The explanation provided so far for the behavioral results assumes that the compensation for
the different angular velocities coming from either the periphery (high angular velocities) or
the center (low angular velocities) of the field of view is not perfect. In particular we assume
that the availability of high angular velocities from near regions directly increases the
estimated speed. On this basis we formulated two hypotheses: when the visibility of the
central region of the field of view is precluded we expect a higher perceived speed, and, on
the other hand, when the peripheral region of the FOV is not visible, we expect a lower
perceived speed. In this experiment we examined whether speeds of forward translations can
be accurately perceived when only limited regions of the FOV are visible.
DSC 2008 Europe Monaco 31 January - 1 February 2008
3.1. Method
3.1.1. Apparatus and stimuli
The same experimental setup as the previous experiment was used. Two buttons on the
steering wheel of the driving interface where used in the experiment.
Because of the perceptual task of the present study, we were not measuring driving
performance, and thus we modeled a short and straight road session instead of the previous
complex road. We kept all the other visual features described in the previous experiment:
traffic divider, road poles and markings. We created three different scenarios (full FOV,
central vision only, peripheral vision only) in which a 40 degrees soft-edge disc-shaped
transparent mask was implemented. In the full FOV scenario (Full) the environment was fully
visible (figure 5a); in the central vision scenario (CV) the mask occluded the peripheral
portion of the scene and only the central 40 degrees were visible (figure 5b); and, finally, in
the peripheral vision scenario (PV) the center of the scene was occluded and only the outer
regions were visible (figure 5c).
Figure 5: The virtual environment with three different transparent masks: the scene entirely visible
(a), the central region visible (b), and the peripheral region visible (c).
3.1.2. Participants
Eight participants with normal or corrected-to-normal vision were recruited. They were paid
and were naïve as to the purpose of the experiment.
3.1.3. Design and procedure
The experiment consisted of a two-interval forced-choice psychophysical task. The
experimental conditions were defined by a pair of visibility masks that were applied to the
first and the second moving scene presented. The four conditions tested were: Full-PV; Full-
CV; PV-CV and CV-PV. For each trial, participants were asked to judge the speed of the two
moving scenes displayed in one of the experimental conditions, and select which scene was
perceived as moving faster. In figure 6 the typical time-course of a trial is described.
DSC 2008 Europe Monaco 31 January - 1 February 2008
Figure 6: Trial example. A fixation cross appears at four degrees below the horizon, followed by
the reference scene, which stays visible for 500 ms. An inter-stimulus-interval (ISI) of 500 ms
separates the two scenes, with the test scene being displayed for another 500 ms. Then, the fixation
cross disappears and a short sentence on the screen invites the participant to provide the answer by
pressing one of two buttons. When a button is pressed, another trial starts.
The four experimental conditions were randomly presented 80 times each for an overall
number of 320 trials and an overall duration of 40 minutes per subject. The reference scene
was moving always at the speed of 90 km/h, whereas the speed of the test scene varied from
trial to trial according to an adaptive procedure. We used a procedure based on a Bayesian
method (Kontsevich & Tyler, 1999) which defines the test speed of the following trial by
optimizing the information that will be gained with the response, taking into account all the
previous knowledge (e.g. tested values and subject answers). The optimization algorithm
takes as initial parameters estimated ranges of the mean speed, the standard deviation, and the
tested speeds, which were determined in a pilot experiment.
3.2. Results
We considered for statistical analysis the Point of Subjective Equality (PSE) [i.e. the speed at
which the test scene was perceived to move as fast as the reference scene]. Results were
pooled between all the participants and the PSE values for the experimental conditions can be
seen in figure 7. Despite small individual differences in the various conditions, the same
general pattern was evidenced in all participants.
At the one-sample t-test analysis we found that the speed in the peripheral vision condition
had to be significantly lowered to be perceived as moving equally fast as the speed of any of
the other reference conditions, either full FOV or central vision. Conversely, the speed in the
central vision condition had to be much higher when compared to the speed of the peripheral
vision condition (figure 7b: PV-CV), but only slightly (and still significantly) increased when
the reference scene was fully visible (figure 7b: Full-CV).
Figure 7: PSE of the perceived speed. The conditions PV-CV and CV-PV have been merged. The
arrow indicates the reference speed (a). The table presents the estimated PSE, the standard errors,
the computed value of the one-sample t-test and the associated significance probability for all the
four original conditions (b).
In figure 7a the data are plotted to show the estimated PSE. The results of the PV-CV and
CV-PV conditions were merged, since both conditions differed only in the order of
DSC 2008 Europe Monaco 31 January - 1 February 2008
presentation. This arrangement allows us to visualize the bias induced by the angular
velocities from the different regions of the visual field. For instance, in the Full-PV condition,
the speed of the test scene with only the peripheral region visible had to be lowered to 60
km/h in order to be perceived as moving as fast as the 90 km/h reference speed. This means
that the perceived speed at the periphery was much higher than that of the entire scene visible.
An ANOVA for repeated measures indicated that the PSEs differed significantly (F(2,14) =
58.03, p < .001) and a corrected post-hoc test (Bonferroni) revealed that all the three
conditions were significantly different one another.
3.3. Discussion
These results clearly show that the perceived speed during forward motion is strongly affected
by the available velocity signals from the scene. According to our predictions, people cannot
accurately compensate for the different velocity signals available when only limited regions
of the FOV are visible. When the central region of the visual field is occluded the speed at the
periphery is perceived as being higher, and conversely, when the information from the
peripheral region is missing the speed at the center is perceived as being lower. The speed
underproduction effect measured in the foggy scenario of the previous experiment seems
therefore to derive from a speed overestimation effect, which is already present at the
perceptual level.
In the condition where the visual information from the central region is always present, the
measured PSE indicates that the speed is nearly correctly estimated, although the missing
peripheral region could in principle bias the result. However, the effect on the perceived speed
is significantly lower when compared to the effect elicited when only the peripheral region is
visible. Therefore, from the comparison between these two conditions, we are allowed to
conclude that the central vision is certainly necessary and seems to be almost sufficient to
correctly estimate the constant speed of forward self-motion. Conversely, the influence of the
peripheral region on the perceived speed seems to be less relevant, although the largest bias
was obtained comparing the central and peripheral regions directly.
Finally, our results indicate that despite a conceivable effect of the periphery, a central field of
view of 40 degrees is already sufficient to provide a nearly correct estimate of the traveling
speed.
4. Conclusions
The experiments of this study were motivated by the previous results concerning the effects of
contrast reduction on the perceived speed. We tried to understand how a reduced visibility
condition could affect the driving behavior with particular emphasis to a naturalistic
implementation of contrast reduction. While driving in realistic fog, like in the first
experiment, the perceived speed is increased due to the reduced visibility of the central area of
the field of view, which leads to a behavioral compensation reflected by the reduction of the
driving speed. The second experiment confirmed the perceptual origin of this effect, showing
a non-optimal compensation for the raw velocity signals. It seems that the speed estimation
process takes into account the velocity signals from both central and peripheral areas of the
visual field. But here we have provided also evidence that the central vision is necessary and
nearly sufficient to provide a correct speed estimate, even in large FOV virtual environments.
The results presented in this paper highlight some critical aspects of the current state of the
DSC 2008 Europe Monaco 31 January - 1 February 2008
visualization in driving simulators. Special attention seems necessary when simulating fog
conditions. Indeed, we showed that the way fog is rendered in virtual environments
dramatically changes the drivers’ behavior and performance. A realistic implementation of a
contrast attenuation function masks the visual field in a sensible way, and consequently biases
the perceived driving speed. On the other hand, we can adapt with relative ease to a uniform
contrast attenuation, which corresponds rather to a “dirty windshield” situation. Finally, the
extent of the FOV and the presence of visual cues at the peripheral regions are important
design questions to be considered for the visual rendering of driving simulators. Our own
results surprisingly showed that an increase of the FOV beyond 40 degrees does not yield
remarkable changes in speed estimation, although the detection of peripheral objects
necessitates a large field of view.
References
Anstis, S. M. (2001). Footsteps and inchworms: Illusions show that contrast affects apparent
speed. Perception , 30 (7), 785-794.
Anstis, S. M. (2003). Moving objects appear to slow down at low contrasts. Neural Networks,
16, 933-938.
Blakemore, M. R., & Snowden, R. J. (2000). Textured backgrounds alter perceived speed.
Vision Research , 40, 629-638.
Brandt, T., Dichgans, J., & Koenig, E. (1973). Differential Effects of Central Versus
Peripheral Vision on Egocentric and Exocentric Motin Perception. Experimental Brain
Research , 16, 476-491.
Casey, S. M., & Lund, A. K. (1987). Three field studies on driver speed adaptation. Human
Factors , 29, 541-550.
Denton, G. G. (1980). The influence of visual pattern on perceived speed. Perception , 9, 393-
402.
Distler, H., Bülthoff, H. H., (1996). Velocity perception in 3-D environments. Perception, 25
ECVP Abstract Supplement.
Dyre, B. P., Schaudt, W. A., & Lew, R. T. (2005). Contrast gradients increase apparent
egospeed while moving through simulated fog [Abstract]. Journal of Vision , 5 (8), 335a.
Evans, L. (1991). Traffic Safety and the Driver. New York: Van Nostrad Reinhold.
Gegenfurtner, K. R., Mayser, H., & Sharpe, L. T. (1999). Seeing movement in the dark.
Nature , 398, 475-476.
Gibson, J. J. (1950). The Perception of the Visual World. Cambridge: Riverside Press.
Green, M. (1983). Contrast detection and direction discrimination of drifting gratings. Vision
Research , 23, 281-289.
Hawken, M. J., Gegenfurtner, K. R., & Tang, C. (1994). Contrast dependence of colour and
luminance motion mechanisms in human vision. Nature , 367, 268270.
Kemeny, A., & Panerai, F. (2003). Evaluating perception in driving simulation experiments.
Trends in Cognitive Sciences , 7 (1), 31-37.
Kontsevich, L. L., & Tyler, C. W. (1999). Bayesian adaptive estimation of psychometric
slope and threshold. Vision Research , 39, 2792-2737.
Koschmieder, H. (1924). Theorie der horizontalen Sichweite. Beitr. Phys. Freien Atmos., 12,
33-53; 171-181.
Moulden, B., Kingdom, F., & Gatley, L. F. (1990). The standard deviation of luminance as a
metric for contrast in random-dot images. Perception , 19 (1), 79-101.
DSC 2008 Europe Monaco 31 January - 1 February 2008
Osaka, N. (1988). Speed estimation through restricted visual field during driving in day and
night: naso-temporal hemifield differences. In A. G. Gale (Ed.), Vision in Vehicles II (pp. 45-
55). Nottingham: Elsevier Science Publishers B.V.
Pretto, P., & Chatziastros, A. (2006). Changes in optic flow and scene contrast affect the
driving speed. DSC 2006 Europe Driving Simulation Conference, (pp. 263-272). Paris.
Pretto, P., & Chatziastros, A. (2006). The role of scene contrast and optic flow on driving
speed. In A. G. Gale (Ed.), Vision in Vehicles XI. Dublin.
Recarte, M. A., & Nunes, L. (1996). Perception of speed in an automobile: estimation and
production. Journal of Experimental Psychology: Applied , 2, 291-304.
Snowden, R. J., Stimpson, N., & Ruddle, R. A. (1998). Speed perception fogs up as visibility
drops. Nature , 392, 450.
Stone, L. S., & Thompson, P. (1992). Human speed perception is contrast dependent. Vision
Res , 32, 1535-1549.
Takeuchi, T., & De, V. (2000). Velocity discrimination in scotopic vision. Vision Research ,
40, 20112024.
Thompson, P. (1982). Perceived rate of movement depends on contrast. Vision Research (22),
377380.
... 1). Yet, several studies suggested that peripheral vision is critical for motion perception, showing that the size of the FoV affects navigation abilities [9,10], postural control111213, speed perception [14,15] and the sensation of self-motion induced by a moving visual stimulus1617181920. For instance, Turano et al. [10] have shown that peripheral visual information is important for establishing and updating an accurate representation of the spatial structure of the environment. ...
... An alternative explanation is that these underestimations merely reflect a range effect [22], i.e., a response bias towards the middle of the range of the possible responses that is commonly observed in subjective assessments. In contrast to other studies that have reported an effect of horizontal FoV size on motion perception14151617181920, we found only a slight reduction of response variability when the size of the FoV was increased, and absolutely no effect on average performance. Our results therefore suggest that when only visual flow information is available, having more than 301 of horizontal field of view may not provide enough useful information to update the internal representation of space. ...
... Banton et al. [26] reported that a 4.8 km/h walking speed was perceived to be about 50% slower than normal during straight-ahead gaze. Pretto et al. [15] found similar results in a driving simulation with reduced visibility conditions. In their experiment, the perceived speed was higher when the central region of the FoV was occluded and lower when the peripheral region of the FoV was occluded. ...
Article
a b s t r a c t Efficient navigation requires a good representation of body position/orientation in the environment and an accurate updating of this representation when the body–environment relationship changes. Such updating is based on the ability to correctly estimate the speed and amplitude of body displacements. Because navigation in virtual worlds often relies on the sole visual information, we investigated to which extent the size of the field of view (FoV) affects two basic aspects of motion perception: (i) the perceived amplitude of rotations about the body vertical axis (Experiment 1) and (ii) the perceived speed of forward translations (Experiment 2). Concerning the perception of rotation amplitude, we found that visual flow information gives rise to inaccurate and very variable estimations, with a systematic underestimation of rotations larger than 301. We also found that the accuracy of the estimations does not depend on the size of the FoV and that horizontal FoVs larger than 301 do not improve the performance. Concerning speed perception, central FoVs smaller than 601 gave rise to an underestimation of the visual speed. On the other hand, occluding the central area leaving only peripheral visual information available induced a systematic over-estimation of visual speed, even when only the central 101 of vision was occluded. Taken together, these results suggest that large FoVs are not required to estimate the amplitude of visual rotations about the vertical axis of the body, whereas central FoVs of at least 601 are advisable when speed perception relies on visual flow information.
... Pretto (2006) reported that drivers tend to reduce their speeds as the scene became foggier and vice versa. The argument is that, drivers perceive higher speed because their central region of visual field is being obstructed during foggy conditions, and as a behavioural compensation, driving speed is reduced (Pretto et al., 2008). Notable examples of fog related accidents in the United States are the multiple vehicle collisions that occurred on I-77 (over Fancy Gap Mountain) and I-64 (over Afton Mountain) which respectively involved 65 and 21 vehicles (Lynn et al., 2002). ...
... that drivers tend to reduce their speeds as the scene became foggier and vice versa. The argument is that, drivers perceive higher speed because their central region of visual field is being obstructed during foggy conditions, and as a behavioural compensation, driving speed is reduced (Pretto et al. 2008). ...
... It can be contested whether this device provides peripheral stimulation. Stimulation of the peripheral visual field may be of particular importance because motion information appears to be transduced most efficiently via peripheral vision (e.g., Brandt et al., 1973;Held et al., 1975;Berthoz et al., 1975;Osaka, 1988;Pretto et al., 2008Pretto et al., , 2009de Winkel et al., 2018). In a comprehensive review, Strasburger et al. (2011) note that in perimetry (i.e., measuring the field of view), the central visual field is considered to be 60 • in diameter. ...
Article
Full-text available
The risk of motion sickness is considerably higher in autonomous vehicles than it is in human-operated vehicles. Their introduction will therefore require systems that mitigate motion sickness. We investigated whether this can be achieved by augmenting the vehicle interior with additional visualizations. Participants were immersed in motion simulations on a moving-base driving simulator, where they were backward-facing passengers of an autonomous vehicle. Using a Head-Mounted Display, they were presented either with a regular view from inside the vehicle, or with augmented views that offered additional cues on the vehicle's present motion or motion 500ms into the future, displayed on the vehicle's interior panels. In contrast to the hypotheses and other recent studies, no difference was found between conditions. The absence of differences between conditions suggests a ceiling effect: providing a regular view may limit motion sickness, but presentation of additional visual information beyond this does not further reduce sickness.
... Dans leur étude, (Snowden, Stimpson et Ruddle 1998) soutiennent qu'en cas de brouillard (qu'ils simulent par une baisse de contraste), la vitesse perçue est plus faible, ce qui explique pourquoi les automobilistes rouleraient plus vite, malgré le danger plus important dû au manque de visibilité. Plus récemment, (Pretto, Vidal et Chatziastros 2008) remettent en cause cette explication. Ils soutiennent que la diminution de vitesse perçue en cas de brouillard ne proviendrait pas de la baisse de contraste, mais du fait que le brouillard réduit la visibilité de façon plus forte pour les éléments éloignés (dans la vision centrale) que pour les éléments proches (dans la vision périphérique). ...
Article
Full-text available
Speed perception is an important task that the driver must perform continuously to control his/her vehicle. It appears that many factors influence this perception: height of the point of view, field of view, realism of the environment, but also realism of audio and proprioceptive rendering. If some high-performance car driving simulators are able to render motion satisfactorily according to all these criteria, it is not always the case. As a result, speed perception is thus often under estimated, leading into producing higher speeds than in real conditions. Perceptual validity is then not good enough to study driver's behavior. To solve this problem, a technique has recently seen the light, which consists of modifying the geometric field of view while keeping the real field of view unchanged. This technique, quantified by a cue defined as the visual scale factor, is studied in-depth in this thesis. We particularly focused on the influence of this visual scale factor on speed perception and determined the existence of a linear relationship between the perceived speed variation and the actual visual scale factor. We have then determined in what extent it is possible to modify dynamically this factor in order to fit at best driving simulation needs. The experiments presented here have all been carried out on the SAAM dynamic driving simulator which has been designed, realized and tuned in the framework of this thesis.
... It is well known that large projection screens with wide field of view (FoV) provide motion cues in the periphery of the visual field that can result in a greater sense of vection (Hettinger & Riccio, 1992; Mohler, Riecke, Thompson, & Bülthoff, 2005), more accurate navigation abilities (Alfano & Michel, 1990), and more accurate perception of self-motion (Pretto, Ogier, Bülthoff, & Bresciani, 2009). For instance, in a driving simulation scenario, a wide FoV provides a better estimation of speed (Jamson, 2000; Pretto, Vidal, & Chatziastros, 2008) while in flight simulation a FoV bigger than 60 degrees helps in the cruise phase (Keller, Schnell, Lemos, Glaab, & Parrish, 2003). However, motion-based simulators often lack the space for large projection screens, and therefore small screens or head mounted displays (HMD) are sometimes used. ...
Article
Full-text available
Different solutions are used on driving simulators to provide visual feedback. In this study, we investigated the influence of projection technology and field of view on drivers performance in a slalom driving task. We tested a head mounted display against a curved projection system on our CyberMotion simulator, based on an anthropomorphic robot arm. The results showed that drivers performed significantly better using the projection screen than the HMD. The FoV and the motion simulation did not have a measurable influence on the performance.
Article
Decision-making skills are essential to successful performance. To train them, coaches frequently use video replays to show their athletes how to best respond when facing specific situations. Recently, it has been shown that presenting the videos in virtual reality (VR) led to enhanced transfer, from the laboratory to the playing field, compared to when the videos were presented on a standard computer screen (CS). Interestingly, although the videos were identical, many participants informally reported that the VR videos felt accelerated compared to those they usually see on television. Here, we tested this claim by having varsity-level basketball players perform a decision-making task concomitantly with a playback speed estimation task. All participants observed the same video clips in the VR and CS conditions, and the video clips were either presented at their normal speed or had been accelerated or decelerated by 10%. Our results revealed that participants perceived the VR videos as significantly faster than the CS videos (mean perceived playback speed of 100.7% ± 2.35% and 94.9% ± 2.24%, respectively). This difference was, however, caused by the CS videos appearing slower than they truly were. Our results indicate that VR videos appear immune to the speed underestimation frequently reported with CS videos.
Article
To determine a reasonable speed limit and ensure traffic safety in a dynamic low-visibility environment with fog, a driving simulator study was conducted. A total of 31 young participants were recruited, and each completed 5 driving simulator trials under varying visibility conditions and speed levels during the daytime. The combined coupling effect of the visibility and driving speed on drivers’ recognition times was explored, and a quantitative model of the recognition time, visibility, and driving speed was established. A determination method and suggested value of a reasonable driving speed limit in dynamic low-visibility conditions were proposed based on the stopping sight distance model. The results show that there are significant differences in the recognition times of drivers under different visibility and speed conditions. The reasonable driving speed limit values in dynamic low-visibility conditions should be based on visibility changes. When the stopping sight distance is 75 m and the visibility is less than 35 m, the speed limit should be 20 km/h. When the visibility is between 35 m and 60 m, the speed limit should be 30 km/h. When the visibility is between 60 m and 140 m, the speed limit should be 50 km/h. When the visibility is greater than 140 m, the speed limit should be 60 km/h. These research results can provide a theoretical reference for the formulation of a VSL in a dynamic low-visibility environment related to fog and reduce crash risk in conditions of inadequate visibility in fog.
Article
Full-text available
Percevoir sa vitesse est une tâche importante que le conducteur doit réaliser constamment pendant la conduite pour contrôler son véhicule. Il apparaît que plusieurs facteurs ont une influence sur la perception de la vitesse : la hauteur du point de vue, le champ de vision, le réalisme de l'environnement, mais aussi le réalisme des restitutions sonores et proprioceptives. Si certains simulateurs de conduite automobile sont suffisamment performants pour restituer le mouvement de façon satisfaisante selon tous ces critères, ce n'est pas forcément le cas de tous. Il en résulte alors une perception de la vitesse souvent sous-estimée, qui se traduit par une vitesse de conduite plus élevée que celle en conduite réelle. La validité perceptive du simulateur de conduite n'est alors plus suffisante pour certaines études de comportement du conducteur. Pour tenter de résoudre ce problème, une technique a récemment vu le jour, consistant à modifier le champ de vision géométrique tout en gardant un champ de vision constant. C'est cette technique, quantifiée par un indice défini comme un facteur d'échelle visuelle, qui est étudiée de façon plus approfondie au cours de cette thèse. Nous avons notamment étudié comment ce facteur d'échelle visuelle influence la perception de la vitesse et avons déterminé la relation linéaire qui existe entre le changement de vitesse perçue et le facteur d'échelle visuelle utilisé. Nous avons ensuite déterminé dans quelle mesure il était possible de modifier dynamiquement ce facteur d'échelle visuelle afin de l'adapter au mieux en fonction des besoins de la simulation. Les expérimentations présentées ici ont toutes été menées sur le simulateur de conduite dynamique SAAM, dont la conception, la réalisation et le réglage ont été effectués dans le cadre de cette thèse.
Article
Efficient navigation requires a good representation of body position/orientation in the environment and an accurate updating of this representation when the body-environment relationship changes. We tested here whether the visual flow alone - i.e., no landmark - can be used to update this representation when the visual scene is rotated, and whether having a limited horizontal field of view (30 or 60 degrees), as it is the case in most virtual reality applications, degrades the performance as compared to a full field of view. Our results show that (i) the visual flow alone does not allow for accurately estimating the amplitude of rotations of the visual scene, notably giving rise to a systematic underestimation of rotations larger than 30 degrees, and (ii) having more than 30 degrees of horizontal field of view does not really improve the performance. Taken together, these results suggest that a 30 degree field of view is enough to (under)estimate the amplitude of visual rotations when only visual flow information is available, and that landmarks should probably be provided if the amplitude of the rotations has to be accurately perceived.
Article
Full-text available
Many horrendous vehicle accidents occur in foggy weather. Drivers know they should slow down because fog reduces visibility, but many still drive too quickly. The `blame' for many such accidents may be due to a perceptual quirk: it appears that drivers think they are driving far more slowly than they actually are in foggy conditions, and therefore increase their speed.
Article
Velocity perception has been investigated in many experiments with stimuli moving in the picture plane (2-D). For example, experiments with sine-wave gratings have shown that high-frequency patterns are perceived as moving faster than low-frequency patterns, and that high-contrast patterns are perceived as moving faster than low-contrast patterns. We investigated the influence of contrast and spatial frequency on perceived velocity in an open-loop driving simulation to determine whether contrast and spatial frequency account for differences in perceived velocity in complex 3-D environments. The simulated scene consisted of a textured road flanked by two meadows. We used road surface textures with different contrast and spatial frequency contents. In a 2AFC paradigm participants were simultaneously presented two driving simulation sequences depicting vehicles moving at different velocities on roads with different surface textures. Participants judged which vehicle was moving faster. Using an adaptive staircase procedure we determined the point of subjective equality for roads with different surface textures. The results show that perceived velocity in a driving simulation does depend on contrast and spatial frequency of the surface texture. Perceived velocity can be increased by increasing the contrast or the relative amount of high spatial frequencies in the surface texture. The relevance of these results for the design of driving simulators is discussed.
Article
In 2 experiments, the authors studied the perception of speed in an automobile as a function of speed, previous acceleration, trajectory, driving experience, and sex of the participants. In Experiment 1, 60 participants estimated the speed at which they traveled by car. In Experiment 2, 30 participants performed an active estimation task with an accelerator to produce a target speed, in addition to the same passive verbal estimation. The results showed a tendency to underestimate speed, and this effect was more pronounced at lower speeds. The predicted overcompensation in the active production task confirmed the general equivalence of both passive and active estimation despite certain differences. Results are discussed from a psychophysical viewpoint, and implications for driving behavior are also considered.
Article
Three field studies of driver speeds were conducted to test the speed adaptation phenomenon and to define the practical implications of its effect. Sites were selected in which the speeds of vehicles previously exposed to high-speed conditions could be contrasted with speeds of vehicles not previously exposed to high speeds. The following conclusions were drawn from the results of this study: Conditions specific to a traffic site, such as legal speed limits, traffic density, and cross-street activity, determine the extent of speed adaptation. Speed perpetuation does not account for observed speed differences between speed-adapted and non-speed-adapted vehicles. The effects observed in the present study were significant but lower than in previous studies, possibly because of overall lower vehicle speeds. These findings indicate that proposals to increase speeds on rural interstates are likely to result in higher speeds on other, connecting roads as well.
Article
When fog is simulated as a global reduction in contrast, apparent egospeed decreases as fog becomes denser (Snowden, Stimpson, and Ruddle, 1998, Nature). However, fog is more realistically modeled as Mei scattering of ambient light, which reduces contrast exponentially as distance increases. In addition to reducing global contrast, this exponential fog model introduces a contrast gradient in depth, which may change the sampling of optical flow to emphasize nearer objects, thereby increasing the rate of global optical flow, which may result in increases in apparent egospeed (Larish & Flach, 1990, JEP:HPP). We examined whether apparent egospeed is affected by this contrast gradient when global contrast is held constant and fog is modeled exponentially. Observers sequentially viewed pairs of 1-3 s computer simulations of observer translation over a textured groundplane. The display pairs consisted of a standard, for which the simulated translational speed and fog density remained fixed throughout the experiment, and a comparison, for which the speed and density each varied independently over five levels. Observers indicated which display produced greater apparent egospeed. For each level of fog density, apparent egospeed and Weber fractions for egospeed discrimination were estimated by fitting 2-parameter sigmoid functions to the proportion of “faster” judgments as functions of translational speed. Results showed that apparent egospeed increased linearly by approximately 5% as the exponential fog density parameter increased 67%. Weber fractions were unaffected (µ=.069). While moving through real fog, this increase in apparent egospeed due to the contrast gradient opposes the decrease in apparent egospeed due to the global reduction in contrast. Hence, a more accurate understanding of how fog density affects apparent egospeed must account for changes in both the contrast gradient and global contrast. Further experiments examining these variables simultaneously will be discussed.
Article
In 2 experiments, the authors studied the perception of speed in an automobile as a function of speed, previous acceleration, trajectory, driving experience, and sex of the participants. Three levels of driving experience were observed: nondrivers, new drivers, and experienced drivers. In Experiment 1, 60 participants verbally estimated the speed at which they traveled by car. In Experiment 2, 30 participants performed an active estimation task with an accelerator to produce a target speed, in addition to the same passive verbal estimation. The results showed a tendency to underestimate speed, and this effect was more pronounced at lower speeds. The predicted overcompensation in the active production task confirmed the general equivalence of both passive and active estimation despite certain differences. Results are discussed from a psychophysical viewpoint, and implications for driving behavior are also considered. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Optokinetic stimuli allow for two perceptual interpretations. The observer may perceive himself as being stationary in a moving surround (egocentric motion perception) or he may experience an illusion of self-motion, so that the actually moving surroundings appear to be stable (exocentric motion perception).Results 1. Circular motion of the entire surroundings (rotating drum) invariably leads to an apparent self-rotation (circularvection: CV), which is indistinguishable from an actual chair rotation. 2. Following stimulus onset, CV begins after a few seconds latency and slowly increases its apparent velocity until its saturation. CV may outlast the visual stimulus by as much as 30 sec. Latencies are independent of stimulus velocity. 3. Even with drum accelerations up to 15/sec2, stationary subjects cannot infer from the lack of vestibular input that only the drum is rotating. 4. With stimulation of the entire visual field or sufficiently large parts of the peripheral retina, the velocity of apparent self-rotation matches stimulus speed up to 90–120/sec. At higher speeds, CV velocity lags behind stimulus speed and results in additional egocentric motion perception. 5. Masking the central visual field by black disks up to 120 in diameter scarcely diminishes CV. Conversely, if peripheral vision is precluded, stimulation of the central field up to 30 in diameter results in exclusive egocentric motion perception of the surround. With a central and peripheral stimulus equivalent in area, the peripheral stimulus predominates CV. 6. Simultaneous presentation of conflicting central and peripheral optokinetic stimuli (i.e., stimuli rotating in opposite directions) has shown that exocentric orientation depends on the peripheral stimulus whereas optokinetic nystagmus and egocentric motion perception rely on the center of the visual field.