Optical Detection of Dangerous Road Conditions - MDPI
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
sensors
Article
Optical Detection of Dangerous Road Conditions
Armando Piccardi * and Lorenzo Colace
NooEL—Nonlinear Optics and OptoElectronics Laboratory and CNIT—Photonic Networks and Technologies
National Lab, Department of Engineering, University ROMA TRE, Via Vito Volterra 62, 00146 Rome, Italy;
lorenzo.colace@uniroma3.it
* Correspondence: armando.piccardi@uniroma3.it
Received: 13 February 2019; Accepted: 12 March 2019; Published: 19 March 2019
Abstract: We demonstrated an optical method to evaluate the state of asphalt due to the presence of
atmospheric agents using the measurement of the polarization/depolarization state of near infrared
radiation. Different sensing geometries were studied to determine the most efficient ones in terms of
performance, reliability and compactness. Our results showed that we could distinguish between a
safe surface and three different dangerous surfaces, demonstrating the reliability and selectivity of
the proposed approach and its suitability for implementing a sensor.
Keywords: optical sensor; scattering from rough surfaces; polarized scattering; polarization
contrast ratio
1. Introduction
The assessment of surface conditions that are exposed to the effects of atmospheric agents is
crucial when dealing with safety, requiring the development of technologies to exploit sensing elements
to assess and reduce the associated risks.
For instance, a number of methods have been developed for ice detection, each focused on a
specific application from aircraft safety to cooling systems or environmental monitoring: these exploit
combined measurements of temperature and humidity [1–3], capacitive [4–6] or piezoelectric [7]
sensors and fiber optic devices [8,9], and some of them are available as commercial systems [10].
Contact sensors for ice detection on roads or other surfaces are moderately reliable, but they have
problems in terms of installation and maintenance, since they need to be mounted directly on the
iced surface. For these reasons, contactless systems based on electrical resistance measurements [11],
reflected radiation at microwaves [12] or optical frequencies [13], and infrared imaging [14,15] have
been proposed.
With reference to road safety, smart systems equipped on both vehicles and the roadside are
desirable when there is the need for a reliable way to distinguish between safe and potentially
dangerous driving situations [16,17].
A number of methods have been developed based on the refractive index difference of asphalt,
water and ice, by measuring light reflection at different wavelengths. Unfortunately, they show limited
reliability due to their low sensitivity, thus preventing selective measurement [18–20]. More recently
it has been demonstrated that more information can be derived from the analysis of the reflected
polarization, which is very sensitive to different environmental conditions [21–24].
Here we propose an optical method to detect different surface conditions based on the
measurement of two orthogonally polarized components of a light beam scattered by the surface. Since
the polarization of light is sensitive to the surface under investigation, its analysis gives information
on the surface itself and it allows to distinguish different conditions.
Sensors 2019, 19, 1360; doi:10.3390/s19061360 www.mdpi.com/journal/sensors
Sensors 2019, 19, 1360 2 of 8
This method was proposed in recent paper where the working principle and some preliminary
results were reported [25]. It was based on the combined measurement of the polarization contrast ratio
and the amplitude of the scattered light, and it focused on the identification of ice on an asphalt surface.
In this work, we started with the original idea and its preliminary results and developed an
optical method to distinguish different surface conditions. Even though, in principle, this method is
not dependent on the type of surface, our work focused on detecting the state of an asphalt surface
and assessing potentially dangerous road conditions.
Using polarimetry measurement, we investigated different sensing configurations in order
to optimize both reliability and selectivity in the detection of asphalt conditions. Furthermore,
we implemented a sensing device and developed an electronic read-out suitable to handle the
optical signals.
2. Materials and Methods
To investigate the different surfaces we employed a diode that emitted at λ = 980 nm. A near
infrared source has several advantages: first of all, it is invisible to human eye and therefore it avoids
the risk of annoying light scattering; it is in a wavelength range where the most common atmospheric
agents (e.g., water and ice) feature a high degree of selectivity [26]; and, additionally, low cost near
infrared optoelectronic devices are widely available. The radiation is modulated with a square wave at
a frequency of about 1 kHz, which optimizes the detection sensitivity and immunity to environmental
(i.e., continuous) light when a narrowband receiver is used.
A lens was used to collimate the radiation and obtain a spot size of about 5 mm on the surface
under investigation. This solution compensates for beam diffraction, allowing maximization of the
light impinging on the detectors. A polarizer can be added to the source to study the sensor response
as a function of incident light polarization. Two further polarizers were used to select the polarization
components of the scattered light parallel and perpendicular to the incident plane and a couple
of photodiodes detected polarized radiation and converted the optical signals into current signals.
The front-end electronics performed a current to voltage conversion by means of trans-impedance
amplifiers (TIA). The signals corresponding to the orthogonal polarizations were detected by a
heterodyne receiver (i.e., a lock-in amplifier) in order to extract and process the information about the
surface condition even in the presence of large noise. The block diagram of the sensor is sketched in
Figure 1a.
The figure of merit that was exploited to distinguish between different surfaces was the
polarization contrast ratio (PC). This parameter is a normalized quantity sensitive to the differences
between the polarization components, which was defined as PC = (ITE − ITM )/(ITE + ITM ), where
ITE/TM is the light intensity corresponding to the two polarization components. In addition, the setup
was equipped with a third photodetector in order to measure the total unpolarized light scattered from
the surface, thus providing a second useful parameter.
Since Fresnel’s coefficients depend on the incidence angle [27], preliminary measurements were
performed to identify the optimal geometric configuration. Thus, we first found the angle for which
the distance among the polarization contrast values corresponding to different surface states was
maximized. Once this angle was found, the polarization contrast was calculated from the converted
voltage values (representing the light intensities) and it was plotted against the total (unpolarized)
scattered light. In this way, the position of the single point on the parameter space (PC vs unpolarized
scattering) is a function of the surface condition [25].
We focused our experiments on asphalt surfaces with the aim of detecting different atmospheric
agents affecting road safety. The investigated samples refer to different asphalt conditions, as indicated
in Figures 2–5 (later in the text). Four kinds of surfaces were tested. The reference condition corresponds
to dry asphalt, which can be identified as a safe condition. Several kinds of asphalts were tested,
differing in both roughness and composition. Though no formal classification of the kinds of asphalt
Sensors 2019, 19, 1360 3 of 8
Sensors 2019, 19, x FOR PEER REVIEW 3 of 8
was performed, we verified (using data in Figures 3 and 5, see Section 3) that all kinds of asphalt
sharedA first potentially
similar dangerouswith
optical properties condition
respectistothe
thepresence
scatteredofpolarized
water on radiation.
the asphalt surface. This has
been Astudied in two different
first potentially dangerous states: wet asphalt
condition and asphalt
is the presence where
of water surface surface.
on the asphalt was completely
This has
covered
been studiedby ainthick layer (i.e.,
two different a few
states: wet millimetres) of water.
asphalt and asphalt whereA the
second
surfacehazardous condition
was completely was
covered
determined by the
by a thick layer presence
(i.e., of ice: we used
a few millimetres) an asphalt
of water. surface
A second that wascondition
hazardous covered by a thin
was ice layer by
determined (1
mm).
the presence of ice: we used an asphalt surface that was covered by a thin ice layer (1 mm).
Thus, we investigated
investigated the the scattered
scattered radiation
radiation from
from four
four different
different states,
states, identified
identified as
as dry,
dry, wet,
wet,
water and
and ice
iceininthe
thefollowing.
following.We We would like to emphasize that it is not within the scope
would like to emphasize that it is not within the scope of this paper of this
paper to investigate
to investigate the optical the properties
optical properties of these
of these kinds kinds ofbut
of surfaces, surfaces,
only tobut only the
validate to validate
approachthe to
approach
efficiently to efficiently
detect detectcondtions.
the surface the surface condtions.
Two major geometries were investigated, as shown in Figure 1. The first was based on Fresnel
reflection with the source and detectors mounted in a θ–2θ θ–2θ configuration
configuration (Figure
(Figure 1b).
1b). Using this
geometry, the horizontal (vertical) polarized reflected radiation was maximized (minimized) at the
Brewster angle (like for a smooth surface) and thus featured a wide range of polarization contrast
ratio with benefits to the system sensitivity. Nevertheless, the system required the source (S) and the
detector (D) to be placed in different positions, increasing increasing the
the space
space needed
needed for for the
the sensor.
sensor.
The alternative geometry measured the polarization components of the backscattered light that
was always present during scattering from a rough surface (Figure 1c). In this case, higher sensitivity
was needed
neededfrom from the the receiver,
receiver, but thebut the resulting
resulting sensor was sensor
muchwasmoremuchcompact.more compact.
In both In both
configurations,
configurations,
the presence of an theadditional
presence of an additional
detector, used to detector,
sense the used to sensescattered
unpolarized the unpolarized
light from scattered light
the direction
from the direction of the surface normal,
of the surface normal, was considered as well. was considered as well.
Figure 1. Block
Blockdiagram
diagramandandset-up.
set-up.(a)(a)Block
Blockdiagram
diagram of of
thethesensor. TheThe
sensor. radiation
radiationfrom thethe
from LED (S)
LED
was collimated
(S) was collimatedby by
a lens
a lens(L)(L)and
andimpinged
impingedon on the
the surface investigation. After
surface under investigation. After being
reflected/scattered,
reflected/scattered,ititwas
wasdetected
detectedbybytwo twophotodiodes
photodiodes (PD1
(PD1 and PD2).
and TwoTwo
PD2). polarizers (P1 (P1
polarizers and and
P2)
selected the horizontal and vertical polarization components. Two trans-impedance
P2) selected the horizontal and vertical polarization components. Two trans-impedance amplifiers amplifiers (TIA)
and
(TIA)the
andlock-in amplifier
the lock-in converted
amplifier the the
converted optical signals
optical into
signals two
into twovoltage
voltagelevels
levelsready
readyforforthe
the signal
signal
processing. (b) θ–2θ
processing. (b) θ–2θconfiguration.
configuration.TheTheincidence
incidence angle
angle θ was
θ was referred
referred to surface
to the the surface normal
normal and aand a
third
third photodiode
photodiode (PD3)
(PD3) was usedwas used tothe
to measure measure the unpolarized
unpolarized scattering. (c)scattering. (c) Backscattering
Backscattering configuration.
configuration.
The set-up wasThe set-up was
analogous analogous
to (b), but bothtophotodetectors
(b), but both collecting
photodetectors collectingscattering
the polarized the polarized
were
scattering
mounted on were
themounted
same sideon ofthe
thesame side of the source.
source.
The experimental set-up was realized with commercial components: the source was a LED940E
(Thorlabs) with a maximum
maximum drivedrive current
current ofof 100 mA
mA for a power of 18 mW. The polarizers were
LPV050 (Thorlabs) with an extinction ratio higher than
(Thorlabs) with an extinction ratio higher than 100000:1
100000:1at the employed
at the employed wavelength andand
wavelength the
detectors
the were
detectors BPW34F
were BPW34F(OSRAM)
(OSRAM) thatthat
worked
workedin the 780–1100
in the 780–1100nmnm spectral
spectralrange and
range andexhibited
exhibiteda
2
amaximum
maximumresponsivity ofof
responsivity 0.70.7
A/W
A/Watatλλ==950
950nm
nmand
andaasensitive
sensitivearea 2.65×× 2.65
areaofof2.65 2.65 mm
mm2.
3. Results
3. Results
In the
In the first
first geometry,
geometry, Figure
Figure 1b,
1b, the
the source
source and
and the
the detectors
detectors were
were mounted on two
mounted on two arms
arms in
in aa
goniometric system to preserve the symmetry of the incident and reflection angles, with the source
goniometric system to preserve the symmetry of the incident and reflection angles, with the source
mounted approximately
mounted approximately 3030 cm
cm from
from the
the surface
surface under
under investigation.
investigation. AA preliminary
preliminary measurement
measurement
involved the evaluation of the polarization contrast as a function of the incident angle [28]. The
results in the case of unpolarized light from the source are shown in Figure 2a, after being averaged
over several acquisitions. Most of the curves corresponding to the different kind of surfaces were
Sensors 2019, 19, 1360 4 of 8
involved
Sensors 2019,the
19, xevaluation of the polarization
FOR PEER REVIEW contrast as a function of the incident angle [28]. The results 4 of 8
in the case of unpolarized light from the source are shown in Figure 2a, after being averaged
superposed,
over severalpreventing
acquisitions. theMost
identification of thecorresponding
of the curves asphalt conditions. to theOnly for θ kind
different = 50° of were the curves
surfaces were
separated, but the polarization contrast range was small. We repeated the same
superposed, preventing the identification of the asphalt conditions. Only for θ = 50 were the curves measurement
◦ using
polarized
separated,light,
but theexploiting the contrast
polarization polarization
rangedependence
was small. We of repeated
the Fresnel coefficients.
the same measurementFigureusing
2b,c
reports on the angular dependence of the scattered light from a source with horizontal
polarized light, exploiting the polarization dependence of the Fresnel coefficients. Figure 2b,c reports and vertical
polarization,
on the angular respectively.
dependenceWhen of the using horizontal
scattered polarization
light from a source with from the source
horizontal and(i.e., perpendicular
vertical to
polarization,
the plane of incidence) the curves exhibited some angles (θ > 50°) where the
respectively. When using horizontal polarization from the source (i.e., perpendicular to the plane of safe condition was
separated
incidence)from all the exhibited
the curves others, but none
some of the(θ“dangerous”
angles > 50◦ ) wherestates were
the safe clearly distinguishable.
condition was separated from On
the other
all the hand,
others, the
but vertical
none of thepolarized
“dangerous”lightstates
(i.e., were
polarization parallel to the plane
clearly distinguishable. On the of other
incidence)
hand,
guaranteed a larger separation between the different conditions, at least around
the vertical polarized light (i.e., polarization parallel to the plane of incidence) guaranteed a larger 50°, where the
polarization contrast value associated with water (for both the ◦water layer and wet
separation between the different conditions, at least around 50 , where the polarization contrast value cases) exhibited a
minimum
associated.with
This water
was due(fortoboth
the the
Brewster angleand
water layer associated
wet cases)to the water surface
exhibited (approximately
a minimum. This was due 53°).
to
The transmitted radiation was minimized, increasing the intensity◦ associated
the Brewster angle associated to the water surface (approximately 53 ). The transmitted radiation was with the reflected
radiation,
minimized, and thus increasing
increasing the associated
the intensity absolute value
with theof the polarization
reflected radiation, contrast
and thus ratio. Thus, we
increasing the
concluded that vertical polarized radiation offered the best performances for
absolute value of the polarization contrast ratio. Thus, we concluded that vertical polarized radiation the realization of a
sensor in this geometry.
offered the best performances for the realization of a sensor in this geometry.
Figure 2.2. θ–2θ
Figure configuration.Polarization
θ–2θconfiguration. Polarization contrast
contrast versus
versus incidence
incidence angle
angle for unpolarized,
for (a) (a) unpolarized,
(b)
(b) horizontally and (c) vertically polarized incident light.
horizontally and (c) vertically polarized incident light.
After setting the vertically polarized source to 50◦ with respect to the surface normal, we measured
After setting the vertically polarized source to 50° with respect to the surface normal, we
the intensities associated with the two polarization components and calculated the polarization contrast
measured the intensities associated with the two polarization components and calculated the
for all the surfaces under investigation. A large number of acquisitions were taken in order to evaluate
polarization contrast for all the surfaces under investigation. A large number of acquisitions were
statistical errors and to simulate the “on field” conditions, including motion and different kinds of
taken in order to evaluate statistical errors and to simulate the “on field” conditions, including
asphalt surfaces.
motion and different kinds of asphalt surfaces.
To evaluate the surface state, we built two parameter planes. The first included the polarization
To evaluate the surface state, we built two parameter planes. The first included the polarization
contrast values versus the total unpolarized scattering, evaluated from the measurement of the
contrast values versus the total unpolarized scattering, evaluated from the measurement of the light
light scattered along the surface normal and collected by the third photodetector. The second was
scattered along the surface normal and collected by the third photodetector. The second was
composed of the polarization contrast versus the sum of the two polarization components. In this
composed of the polarization contrast versus the sum of the two polarization components. In this
way, each surface condition could be defined by a sector of the parameter plane, as the values of
way, each surface condition could be defined by a sector of the parameter plane, as the values of the
the experimental measurements corresponding to the different surfaces generated a group of points
experimental measurements corresponding to the different surfaces generated a group of points
(with
(with thethepoint
pointdispersion
dispersionbeing
beingdue
duetoto measurement
measurement uncertainty,
uncertainty, thethe different
different kinds
kinds of asphalt,
of asphalt, or
or slightly different surface conditions, like water or ice layer thickness), which should be
slightly different surface conditions, like water or ice layer thickness), which should be included in included in
the corresponding part of the
the corresponding part of the plane.plane.
Figure 33 shows
Figure shows the
the graphs
graphs obtained
obtained for
for aa vertical
vertical polarized
polarizedsource
sourceand
andincidence
incidenceangle
angleof 50◦ .
of50°.
The former
The former case,
case, PC
PCversus
versusunpolarized
unpolarizedscattering
scattering(Figure
(Figure3a), exhibited
3a), a problem
exhibited withwith
a problem the dry
the (i.e., safe)
dry (i.e.,
and iced (i.e., dangerous) conditions, where the two groups of points were
safe) and iced (i.e., dangerous) conditions, where the two groups of points were partially partially overlapping.
Thus, the choice
overlapping. of geometry
Thus, the choice andofmeasured
geometryparameters
and measureddid notparameters
allow selective
did measurement for the
not allow selective
considered surface
measurement states.
for the Conversely,
considered using
surface the sum
states. of the orthogonal
Conversely, using thepolarizations
sum of theallowed us to
orthogonal
polarizations allowed us to discriminate among all four different states even if the voltage signal
levels were slightly lower (thus, the required sensitivity was higher), as shown in Figure 3b. Even if
the water and wet conditions were very close, they almost correspond to the same kind of dangerous
Sensors 2019, 19, 1360 5 of 8
discriminate among all four different states even if the voltage signal levels were slightly lower (thus,
Sensors 2019, 19,
the required
Sensors 2019, 19, xx FOR PEER
PEER REVIEW
sensitivity
FOR REVIEW
was higher),
as shown in Figure 3b. Even if the water and wet conditions 55 of
of 88
were very close, they almost correspond to the same kind of dangerous conditions, thus the risk of
conditions, thus
conditions, thus the
the risk
risk of
of confusing
confusing the twotwo surfaces
surfaces is
is not
not critical
critical and
and it
it would
would not
not affect
affect the
the
confusing the two surfaces is not criticalthe
and it would not affect the performance of the sensor in an
performance
performance of the sensor in an
of theinvestigation. application-based investigation.
sensor in an application-based investigation.
application-based
Figure
Figure 3.
Figure 3. Mirror-like
Mirror-likeconfiguration.
Mirror-like configuration.Polarization
configuration. Polarizationcontrast
Polarization contrast
contrast versus
versus
versus (a)
(a)(a) the
thethe total
total scattering
scattering
total and
andand
scattering (b)sum
(b) the
(b) the
the
sum
of theof
twothe two orthogonal
orthogonal polarized
polarized components
components
sum of the two orthogonal polarized components (V and(V
V TE and VTM, the voltage corresponding to
(VTE , the voltage corresponding to horizontal
TE TMand VTM, the voltage corresponding to
horizontal
and vertical
horizontal and vertical polarization,
polarization,
and vertical polarization, respectively).
respectively).
respectively).
The polarization
The polarization contrast
polarization contrast was
contrast was close
was close to
close to zero for
to zero for both
both the
the dry
dry asphalt
asphalt and
and the
the ice
ice cases,
cases, which
which
indicates that
indicates that the
the scattered
scattered light
light was
was almost
almost unpolarized,
unpolarized, probably
probably duedue to
to the
the depolarization
depolarization effects
effects
from the
the high
high surface
surface roughness
roughness inin both
both cases.
cases. TheThe presence
presence of of water
water enhanced
enhanced
from the high surface roughness in both cases. The presence of water enhanced light polarization,lightlight polarization,
polarization, even
even if
if the if
even the
groups groups
of points
the groups of points corresponding
ofcorresponding to
to the wettoand
points corresponding the wet
thewater and
wet layer water layer
were quite
and water were
layerdispersed.quite
were quite dispersed.
Nevertheless,
dispersed.
Nevertheless,
this geometry this
allows geometry
us to allows
discriminate us
amongto discriminate
some among
potentially some
dangerous
Nevertheless, this geometry allows us to discriminate among some potentially dangerous potentially
conditions. dangerous
Furthermore,
conditions. Furthermore,
the third photodiode
conditions. Furthermore, the
is not
the third photodiode
photodiode is
necessary.
third is not
not necessary.
necessary.
We performed
We performed the the same
same kind
kind ofof measurement
measurement procedure
measurement procedure
procedure inin the
the backscattering
backscattering configuration.
backscattering configuration.
configuration.
As for
for the
the former
former geometry,
geometry, we we
first first verified
verified which which
angle angle maximized
maximized the
As for the former geometry, we first verified which angle maximized the separation between the
separationseparation
between between
different
different
conditions,
different conditions,
as sketched
conditions, as sketched
asinsketched
Figure 4, inin Figure
for Figure 4, for
the unpolarized the unpolarized
4, for the(Figure (Figure
4a) and (Figure
unpolarized polarized 4a) and polarized
(horizontally
4a) and
and polarized
(horizontally
vertically, and
Figure vertically,
4b,c, Figures
respectively) 4b,c,
incidentrespectively)
radiation. incident
(horizontally and vertically, Figures 4b,c, respectively) incident radiation. radiation.
Figure 4.4.Backscattering
Figure Backscatteringconfiguration. Polarization
configuration. contrast
Polarization versus incidence
contrast angle for (a)
versus incidence
incidence unpolarized,
angle for (a)
(a)
Figure 4. Backscattering configuration. Polarization contrast versus angle for
(b) horizontally
unpolarized, (b) and (c) vertically
horizontally and polarized
(c) incident
vertically light.incident light.
polarized
unpolarized, (b) horizontally and (c) vertically polarized incident light.
It can
It can bebe noted
notedthat
thatthe
thepolarization
polarizationcontrast
contrast was
waslower
lowerwith respect
with to the
respect previous
to the
the case.case.
previous ThisThis
was
It can be noted that the polarization contrast was lower with respect to previous case. This
expected
was expectedconsidering
expected considering the lower
considering the signal
the lower level
lower signal due
signal level to different
level due
due to geometry.
to different Nevertheless,
different geometry. here
geometry. Nevertheless, the
Nevertheless, herefour road
here the
the
was
conditions
four road often exhibited
conditions often quite different
exhibited quitevalues. In the
different unpolarized
values. In the case the whole
unpolarized case angle
the range below
whole angle
four road conditions often exhibited quite different values. In the unpolarized case the whole angle
θ = 40◦below
range couldθbe= used, evenbe
40° could
could if the dry
used, condition
even if the wascondition
the dry
dry always close wasto the iceclose
always curve. toWhen iceexamining
the ice curve. When
Whenthe
range below θ = 40° be used, even if condition was always close to the curve.
horizontal
examining the polarized case,
the horizontal we can
horizontal polarized easily
polarized case, see
case, we that
we canthe curves
can easily
easily see were
see thatsuperposed
that the
the curves for
curves were almost all
were superposed the angles
superposed for for
examining
almost all
almost all the
the angles
angles (except
(except for for θ
θ == 65°,
65°, where
where drydry and
and iceice conditions
conditions were
were still
still too
too close
close to
to ensure
ensure
reliable detection), thus this case was also not taken into consideration. The vertical
reliable detection), thus this case was also not taken into consideration. The vertical polarized light polarized light
case provided much better results, exhibiting several angles where
case provided much better results, exhibiting several angles where all the curves were well all the curves were well
separated. In
separated. In particular,
particular, at
at θ θ == 20°
20° the
the graph
graph shows
shows almost
almost equal
equal distances
distances among
among the the curves
curves in in aaSensors 2019, 19, 1360 6 of 8
(except for θ = 65◦ , where dry and ice conditions were still too close to ensure reliable detection), thus
this case
Sensors was
2019, 19, xalso
FORnot
PEERtaken into consideration. The vertical polarized light case provided much better
REVIEW 6 of 8
results, exhibiting several angles where all the curves were well separated. In particular, at θ = 20◦
quite largeshows
the graph interval of about
almost equal0.12 for polarization
distances among the contrast.
curvesAs inaafurther advantage
quite large intervalwith respect
of about to the
0.12 for
unpolarized case, the As
polarization contrast. dryacondition had the highest
further advantage contrast
with respect value,
to the and it was
unpolarized well
case, theseparated from
dry condition
all
hadthethe other
highest potentially dangerous
contrast value, and it situations. In addition,
was well separated frominallthis
the case,
other the verticaldangerous
potentially polarized
radiation
situations. was considered
In addition, in thistocase,
be the
thevertical
best choice to radiation
polarized discriminate among theto different
was considered be the bestasphalt
choice
conditions.
to discriminate among the different asphalt conditions.
Thus,
Thus, onceonce thethe incident
incident angle
angle was set to
was set to about
about 20 ◦ , we
20°, we measured
measured thethe two
two polarized
polarized scattering
scattering
components.
components. The The results
resultsare
areshown
shownininFigure
Figure5,5,with
withthethe polarization
polarization contrast
contrast as aasfunction
a function of both
of both the
the total scattering (Figure 5a) and the sum of the signals corresponding to the
total scattering (Figure 5a) and the sum of the signals corresponding to the two orthogonal polarizations two orthogonal
polarizations
(Figure 5b). (Figure 5b).
Backscattering configuration. Polarization contrast versus (a) the total scattering and (b) the
Figure 5. Backscattering
sum of the two orthogonal polarized components.
Remarkably, the
Remarkably, the two
two graphs
graphs were
were very
very similar,
similar, which
which made
made the
the third
third photodiode
photodiodeunnecessary.
unnecessary.
In other words, the backscattered light contained all the information needed, allowing usreduce
In other words, the backscattered light contained all the information needed, allowing us to the
to reduce
sensor complexity.
the sensor complexity.
The dry
The dry asphalt
asphalthad hadthe
thehighest value
highest of both
value contrast
of both and total
contrast scattering,
and total thus defining
scattering, a region
thus defining a
of safe condition in the parameter space, while the polarization contrast easily allowed
region of safe condition in the parameter space, while the polarization contrast easily allowed the the detection
of the presence
detection of theof ice. Theof
presence water layerwater
ice. The showedlayertheshowed
lower value of (back)scattering
the lower due to the smooth
value of (back)scattering due to
surface, while the group of points corresponding to the wet condition was in
the smooth surface, while the group of points corresponding to the wet condition was in the middlethe middle between
water
betweenandwater
dry, as
andexpected. The results
dry, as expected. forresults
The each group
for eachof points
groupwere well were
of points separated
well from the others
separated from
with reasonable dispersion occurring, which allowed us to divide the parameter
the others with reasonable dispersion occurring, which allowed us to divide the parameter space space into four parts
that four
into wereparts
associated to the
that were different to
associated asphalt conditions.
the different Forconditions.
asphalt example, PC = example,
For 0.05 would PCdivide
= 0.05the dry
would
asphalt and the ice states.
divide the dry asphalt and the ice states.
Hence, the
Hence, thegeometry
geometry thatthat
employed the backscattered
employed radiation radiation
the backscattered guaranteedguaranteed
the best performance.
the best
Moreover, employing only two photodiodes mounted on the same
performance. Moreover, employing only two photodiodes mounted on the same side side of the source allowed for the
of the source
most compact
allowed for theconfiguration
most compact for configuration
the practical realization of sensors
for the practical based onof
realization this method.
sensors based on this
method.
4. Discussion
4. Discussion
The presented results demonstrate the possibility of implementing an optical sensor to distinguish
several asphalt conditions. The information on the state is given by the position of the measured
The presented results demonstrate the possibility of implementing an optical sensor to
values on the parameter plane: for instance, Figure 4 suggests that the plane can be divided into four
distinguish several asphalt conditions. The information on the state is given by the position of the
parts corresponding to the four different states. The edges of the four parts define threshold values
measured values on the parameter plane: for instance, Figure 4 suggests that the plane can be
of the polarization contrast and the sum of the components, which could be represented by straight
divided into four parts corresponding to the four different states. The edges of the four parts define
lines. In other words, our results can be used as a calibration of the system. A sensor employing this
threshold values of the polarization contrast and the sum of the components, which could be
method will measure the suitable quantities by placing the corresponding point on the parameter
represented by straight lines. In other words, our results can be used as a calibration of the system. A
sensor employing this method will measure the suitable quantities by placing the corresponding
point on the parameter plane. The surface state will be identified by comparing the position of the
measured point with the calculated thresholds.
Future developments of our work will encompass the increased number of detected conditions,
including a study of the perturbations coming from disturbances such as the presence of dust orSensors 2019, 19, 1360 7 of 8
plane. The surface state will be identified by comparing the position of the measured point with the
calculated thresholds.
Future developments of our work will encompass the increased number of detected conditions,
including a study of the perturbations coming from disturbances such as the presence of dust or water
particles in the air, mixed surface conditions, or fast motion. Moreover, the implementation of an
algorithm to define the threshold values has to be performed. This should include the definition of
the boundaries for each zone, for instance by the statistical analysis of the groups of points over a
huge number of acquisitions. Finally, the development of electronics able to automatically process the
measured data and provide an alert signal linked to the state of the surface will be required.
5. Conclusions
We proposed an optical method for the detection of dangerous road conditions based on the
measurement of the polarization state of the scattered radiation from an asphalt surface by two
photodiodes that detected two orthogonal polarized light components. Two main geometries were
investigated. The first measured the reflected light, while the second measured the backscattered
radiation. In both configurations we also evaluated the need for a third photodiode that measured
the total unpolarized scattering. The backscattering geometry was the most reliable, allowing for
discrimination among dry, wet, water and ice surfaces, and thus distinguishing the safe condition
from the other three different danger levels. In addition, in this case the information from the third
photodiode were redundant, allowing a compact sensor geometry.
Thus, we demonstrated the feasibility of a contactless sensor that featured reliability, compactness
and was low cost due to the usage of commonly-used commercial components. We stress that the
method used here is not dependent on the surface under investigation, and thus it is not limited to the
automotive sector. In fact, the reference surface can change from case to case, e.g., a metallic surface
(rather than asphalt) in the case of ice detection on wind turbines or airplane wings. Thus, the same
approach can potentially be applied to any case where different surface conditions must be identified.
Author Contributions: Conceptualization, L.C.; methodology, investigation, A.P. and L.C.; validation, formal
analysis, data curation, A.P.; resources, L.C.; writing—draft preparation, review and editing, A.P. and L.C.;
visualization, A.P. and L.C.; supervision, project administration, funding acquisition, L.C.
Funding: This work was supported by ACTPHAST (Access CenTer for PHotonics InnovAtion Solutions and
Technology Support), grant number P2016-29 (SODARC—Sensor for the Optical detection of DAngerous Road
Conditions).
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Ikiades, A.; Howard, G.; Armstrong, D.J.; Konstantaki, M.; Crossley, S. Measurement of optical diffusion
properties of ice for direct detection ice accretion sensors. Sens. Actuator A 2007, 140, 24–31. [CrossRef]
2. Wollenweber, G.C. Icing Condition Detection Using Instantaneous Humidity Sensing. U.S. Patent App.
15,275,013, 29 March 2018.
3. Hydrologicalusa. Available online: https://www.hydrologicalusa.com/products/snow-ice-measurement/
ice-detection-sensor/ (accessed on 6 February 2019).
4. Zhi, X.; Cho, H.C.; Wang, B.; Ahn, C.H.; Moon, H.S.; Go, J.S. Development of a Capacitive Ice Sensor to
Measure Ice Growth in Real Time. Sensors 2015, 15, 6688–6698. [CrossRef] [PubMed]
5. Troiano, A.; Pasero, E.; Mesin, L. Remote Monitoring of Ice Formation over a Runway Surface. IEEE Trans.
Instr. Meas. 2010, 60, 1091–1101. [CrossRef]
6. Flatscher, M.; Neumayer, M.; Bretterklieber, T. Maintaining critical infrastructure under cold climate
conditions: A versatile sensing and heating concept. Sens. Actuators A Phys. 2017, 267, 538–546. [CrossRef]
7. Roy, S.; De Anna, R.G.; Mehregani, M.; Zakar, E. A capacitive ice detection microsensor. Sens. Mater. 2000,
12, 1–14.Sensors 2019, 19, 1360 8 of 8
8. Ikiades, A. Direct ice detection based on fiber optic sensor architecture. Appl. Phys. Lett. 2007, 91, 104104.
[CrossRef]
9. Amiropoulos, K.; Spasopoulos, D.; Ikiades, A. Fiber optic sensor for ice detection on aerodynamic surfaces
using plastic optic fiber tapers. In Optical Sensors; OSA Technical Digest; Optical Society of America:
Washington, DC, USA, 2018.
10. Vaisala DRS511. Road & Runway Sensor. Available online: http://www.vaisala.com (accessed on 6
February 2019).
11. Tabatabai, H.; Aljuboori, M. A Novel Concrete-Based Sensor for Detection of Ice and Water on Roads and
Bridges. Sensors 2017, 17, 2912. [CrossRef] [PubMed]
12. Finkele, R. Optical road-ice detector operating in the near infrared. Electron. Lett. 1997, 33, 1153–1154.
[CrossRef]
13. Ogura, T.; Kageyama, I.; Nasukawa, K.; Miyashita, Y.; Kitagawa, H.; Imada, Y. Study on a road surface
sensing system for snow and ice road. JSAE Rev. 2002, 23, 333–339. [CrossRef]
14. Muñoz, C.; Márquez, F.; Tomás, J. Ice detection using thermal infrared radiometry on wind turbine blades.
Measurement 2016, 93, 157–163. [CrossRef]
15. Abdel-Moati, H.; Morris, J.; Zeng, Y.; Wesley Corie, M.; Garas Yanni, V. Near field ice detection using infrared
based optical imaging technology. Opt. Laser Technol. 2018, 99, 402–410. [CrossRef]
16. EPoSS Roadmap Automated Driving 2014. Available online: https://www.smart-systems-integration.org/
publication/eposs-roadmap-automated-driving-2014 (accessed on 14 March 2019).
17. Guerrero-Ibáñez, J.; Zeadally, S.; Contreras-Castillo, J. Sensor Technologies for Intelligent Transportation
Systems. Sensors 2018, 18, 1212. [CrossRef] [PubMed]
18. Giles, D.B.; Kemp, D.R. Method and Apparatus for Detecting Ice and Packed Snow. U.S. Patent No. 5,818,339,
6 October 1998.
19. Schmitt, K.; Schaube, W. Method for Determining the Condition of a Roadway Surface. U.S. Patent No.
5,218,206, 8 June 1993.
20. Casselgren, J.; Sjödahl, M.; LeBlanc, J. Angular spectral response from covered asphalt. Appl. Opt. 2007, 46,
4277–4288. [CrossRef] [PubMed]
21. Lu, Y.; Higgins-Luthman, M.J. Black Ice Detection and Warning System. U.S. Patent Appl. No. 2008/0129541
A1, 5 June 2008.
22. Stern, H.; Schaefer, J.; Maali, F. System for Detecting Ice or Snow on Surface Which Specularly Reflects Light.
U.S. Patent No. 6,069,565, 30 May 2000.
23. Zhao, D. Inflight Ice Detection System. U.S. Patent 7,370,525, 13 May 2008.
24. Casselgren, J.; Sjödahl, M. Polarization resolved classification of winter road condition in the near-infrared
region. Appl. Opt. 2012, 51, 3036–3045. [CrossRef] [PubMed]
25. Colace, L.; Santoni, F.; Assanto, G. A near-infrared optoelectronic approach to detection of road conditions.
Opt. Laser Eng. 2013, 51, 633–667. [CrossRef]
26. Kou, L.; Labrie, D.; Chylek, P. Refractive indices of water and ice in the 0.65- to 2.5-µm spectral range.
Appl. Opt. 1993, 32, 3531–3540. [CrossRef] [PubMed]
27. O’Donnell, K.A.; Knotts, M.E. Polarization dependence of scattering from one-dimensional rough surfaces.
J. Opt. Soc. Am. A 1991, 8, 1126–1131. [CrossRef]
28. Videen, G.; Hsu, J.; Bickel, W.S.; Wolfe, W.L. Polarized light scattered from rough surfaces. J. Opt. Soc. Am. A
1992, 9, 1111–1118. [CrossRef]
© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).You can also read
