See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/306268424

SiDIVS: Simple Detection of Inductive Vehicle Signatures with a Multiplex

Resonant Sensor

Article in Sensors · August 2016

DOI: 10.3390/s16081309

CITATIONS READS

10 7,276

4 authors, including:

Jose Juan Lamas Paula-Maria Castro-Castro

Universidade da Coruña Universidade da Coruña
16 PUBLICATIONS 58 CITATIONS 127 PUBLICATIONS 442 CITATIONS

SEE PROFILE SEE PROFILE

Adriana Dapena
Universidade da Coruña
132 PUBLICATIONS 698 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jose Juan Lamas on 27 August 2016.

The user has requested enhancement of the downloaded file.

 sensors
Article
SiDIVS: Simple Detection of Inductive Vehicle
Signatures with a Multiplex Resonant Sensor
José J. Lamas-Seco † , Paula M. Castro *,† , Adriana Dapena † and Francisco J. Vazquez-Araujo †
Department of Electronics and Systems, University of A Coruña, 15071 A Coruña, Spain;
[email protected] (J.J.L.-S.); [email protected] (A.D.); [email protected] (F.J.V.-A.)
* Correspondence: [email protected]; Tel.: +34-981-167-000 (ext. 6-024); Fax: +34-981-167-160
† These authors contributed equally to this work.

Academic Editor: Vittorio M. N. Passaro

Received: 29 June 2016; Accepted: 15 August 2016; Published: 17 August 2016

Abstract: This work provides a system capable of obtaining simultaneous inductive signatures of
vehicles traveling on a roadway with minimal cost. Based on Time-Division Multiplexing (TDM) with
multiple oscillators, one for each inductive loop, the proposed system detects the presence of vehicles
by means of a shift in the oscillation period of the selected loop and registers the signature of the
detected vehicles by measuring the duration of a fixed number of oscillator pulses. In order to test
the system in an actual environment, we implement a prototype that we denote as SiDIVS (Simple
Detection of Inductive Vehicle Signatures) and acquire different vehicle inductive signatures under
real scenarios. We also test the robustness of the detector by simulating the effect of noise on the
signature acquisition.

Keywords: data acquisition; inductive loop detector; instrumentation and measurements; intelligent
transportation systems; multiplex systems; vehicle inductive signature

1. Introduction
One of the most important aspects of Intelligent Transportation Systems (ITS) is vehicle traffic
monitoring, essentially those applications whose aim is to count the number of vehicles on a roadway
or to know their speed, occupancy, or structural characteristics like density and type.
The sensors used for these applications can be classified, according to their location in the road,
as in-roadway and over-roadway. In general, in-roadway sensors may have problems of installation
and maintenance, whereas over-roadway sensors suffer from blocking of Line-Of-Sight (LOS) by
weather conditions or obstacles. However, the most frequently-used classification of sensors is done
according to the existence or not of an external energy source. Thus, we can talk of active and passive
sensors [1–3], respectively. Passive sensors measure reflected radiation that has been emitted from
the surrounding environmental elements. Some examples of passive sensors are image processing,
acoustic, seismic or magnetic sensors. Active sensors require their own source of excitation and
measure the reflected energy. Radar, laser, infrared, ultrasound and emerging technologies-based
sensors are some examples that lie in this category.
Since their introduction in the 1960s, the active sensors known as Inductive Loop Detectors (ILD)
are the more commonly used sensors in traffic management systems. This type of sensor has the
advantages of being a highly developed technology; of a simple operation, unaffected by environmental
conditions; and of a low installation cost. Although they have the problem of complex replacement in
case of breakage, with the inevitable interruption of traffic, and need regular calibration, their flexible
design can adapt to a large variety of applications, providing better accuracy in traffic classification
than other commonly used techniques [1]. Thus, ILDs have been widely used for tasks such as vehicle
classification [4–8], vehicle re-identification [9–12], and speed estimation with a single loop [13,14].

Sensors 2016, 16, 1309; doi:10.3390/s16081309 www.mdpi.com/journal/sensors

Sensors 2016, 16, 1309 2 of 20

Several works set the basis for the theoretical study of ILDs. In [15,16], an approximate model for
an ILD is shown, and an equivalent circuit model is detailed in [16]. A detailed study on the sensitivity
of an inductive loop and the response time to changes in inductance of different digital detectors is
shown in [1]. The work in [17] addressed the sensitivity problems in dual loops and proposed software
and hardware implementation solutions to identify and correct them. In [18], three-dimensional
maps illustrating the sensitivity of various inductive loops located under the asphalt surface of roads
are obtained.
These works have led to different implementations of ILDs. In the US4680717A patent [19],
a multiplex system for vehicular traffic detection with a single oscillator is presented. A multiple
vehicle detection system incorporating low-cost oscillators and an algorithm to calibrate the device
following changes in environmental conditions is proposed in [20]. However, although both works
allow the detection of the presence of vehicles, neither of them register their inductive signatures.
In [21], a system to detect multiple vehicles is also developed, but it requires multichannel acquisition
of analogue signals, which makes it overly complex. Furthermore, due to the function not being
fully multiplexed and to the use of the same frequencies in near loops, such development causes
significant interferences between channels (also known as crosstalk). The work in [22] also presents
a very complex hardware with a detector that obtains inductive vehicle signatures by measuring
changes in coil impedance, separating its real part (R) from its imaginary part (X). It uses an analogue
hardware, which integrates a self-balanced bridge when no vehicles are present, and two synchronous
demodulators for obtaining the R and X signatures corresponding to each vehicle.
The US691182982 patent [23] presents equipment to produce a vehicle inductive signature by
means of a change in inductance induced in the vehicle-loop when this vehicle passes over the road
loop. Although this amplitude detector allows multiple vehicle detection, it requires complex and
expensive analogue processing and signal acquisition circuits.
Taking into account the drawbacks identified in the aforementioned works, in this paper, we
propose a multiplex system for the Simple Detection of Inductive Vehicle Signatures (SiDIVS).
Our proposal implies a fully multiplexed system that avoids the interference between loops (also
called inter-loop interference) thanks to a very simple and almost fully automatic digital measurement
process. Therefore, it does not require the use of complex and expensive analogue processing circuits
or of analogue signal acquisition methods.
The paper is organized as follows. Section 2 presents a brief theoretical study of ILDs. The models
used to study the impact of noise on amplitude and resonant detectors are presented in Section 3.
Section 4 shows the practical implementation of the SiDIVS prototype using a multiplex detector
with eight channels. Section 5 explains the experimental measurements performed to evaluate
the performance of such digital detectors with the inductive signatures captured by our prototype.
Section 6 presents an analysis of the impact of the noise on these digital detectors. In addition, some
examples of real inductive signatures collected from different vehicles are included. Finally, Section 7
is devoted to the conclusions.

2. Inductive Loop Detectors

Figure 1 depicts the elements of an ILD. It consists of one or more coils with one or more turns
(usually three to five) embedded in the road pavement; isolated cables for the connection from the
coils to the control cabinet; and the electronic equipment (i.e., the detector) inside the cabinet.
The sensitivity S of an inductive loop is a measurement of its ability to detect small changes in
inductance and is defined as the ratio between the change in inductance due to passing traffic and
the initial inductance (i.e., the inductance when a vehicle is not present). Thus, we can express the
sensitivity as
∆L Lnv − Lv
S= = , (1)
Lnv Lnv
Sensors 2016, 16, 1309 3 of 20

where Lnv is the initial inductance when no vehicle is present, and Lv is the inductance when a vehicle
is present, so that the change in inductance is defined as ∆L = Lnv − Lv .

Electronic detector Cabinet

Inductive loop

Wire leads

Figure 1. Elements of an inductive loop detector.

Modern inductive detectors of vehicle presence are digital because they provide more reliable,
accurate, and precise measurements than analogue detectors. Although there exist ILDs measuring
other types of variations in the coil, like impedance [22], currently, the majority of ILDs indirectly
measure variations in inductance as indicated in Equation (1). These variations are caused by the
presence of a vehicle in the detection area of the inductive loop, which produces a decrease of
inductance. Two methods are used to measure such variations: one based on measuring the frequency
or period changes of an oscillator resonant circuit, and one based on measuring the voltage amplitude
changes of an RLC circuit operating at a fixed frequency, known as resonant ILDs and amplitude ILDs,
respectively. We briefly describe both ILD types in the following subsections.

2.1. Resonant ILDs

Resonant ILDs are based on the measurement of changes of oscillation frequency or period.
The oscillator frequency is controlled by a parallel resonant circuit, also called tank circuit, which is
constituted by a non-ideal loop with inductance L in serial with a resistor R, and this serial set
connected in parallel with a capacitance C placed in the detector. The complex impedance of this
circuit is given by:

R 1 − w2 LC + w2 RLC + j wL 1 − w2 LC − wR2 C




Z( jw) = 2
. (2)
(1 − w2 LC ) − w2 R2 C2

Since for a resonant circuit it verifies Im[Z( jw)] = 0, we have that w0 L(1 − w02 LC ) − w0 R2 C = 0,
which gives us the resonant angular frequency w0
r
L − R2 C
w0 = . (3)
CL2

Since our detector has R ≈ 1 Ω, L ≥ 50 µH, and C ≤ 100 nF, it verifies L  R2 C, and we can
approximate w0 by: r
∼ L 1
w0 = = √ , (4)
CL2 LC
Sensors 2016, 16, 1309 4 of 20

so that the oscillation frequency f 0 is given by:

w0 ∼ 1
f0 = = √ . (5)
2π 2π LC
−1/2 ,
√ −1to note that the oscillation frequency depends on the inductance as f 0 = kL
It is important
with k = (2π C ) , and the frequency change is given by ∆ f = f v − f nv , f v being the oscillation
frequency with vehicle presence, and f nv the oscillation frequency without its presence. Thus, we have:
− 1 − 1
∆L ∆L
 
−1 − 12 −1 2 2
fv = kLv 2 = k ( Lnv − ∆L) = kLnv2 1− = f nv 1− , (6)
Lnv Lnv

with f nv = kL− 1/2

nv , and
∆f f v − f nv fv 1
= = −1 = q − 1. (7)
f nv f nv f nv 1− ∆L
Lnv

Since ∆L/Lnv is very small, ∆ f / f nv can be approximated by the first two terms of the Taylor
series, i.e.,
∆f 1 ∆L 1
≈ = S, (8)
f nv 2 Lnv 2
where S is the sensitivity of the inductive loop (see Equation (1)). This sensitivity can thus be
approximated by:
∆L ∆f f v − f nv Tnv − Tv ∆T
S= ≈2 =2 =2 =2 , (9)
Lnv f nv f nv Tv Tv
where Tv = 1/ f v is the period of oscillation if a vehicle is over the coil, and Tnv = 1/ f nv denotes
the period of oscillation otherwise. Experimental results have shown that the loop sensitivity S is
extremely repeatable for fixed sizes and geometries of both the loop and the vehicle and for a fixed
distance between them, as can be verified from Equations (1) and (9).
Detectors whose operation is based on period changes (i.e., based on period shifts ∆T), known as
type III or type IV, in which period shifts or relative period shifts are measured, respectively, present a
measurement time that is short enough for their use in applications of inductive signature capturing,
although the characteristics of the oscillation loop have influence on the threshold sensitivity.
Type III detectors based on period shift use a reference clock signal whose frequency is of several
MHz, typically between 20 and 1000 times greater than the oscillation frequency of the inductive loop
we are interested in measuring. The period of the oscillation signal is calculated as the number of
cycles N of the reference clock signal in m cycles of the oscillation signal. When a vehicle stops or
passes over the loop, the oscillation frequency increases; thus, the period (and thus the number of
cycles N) decreases. The counter of clock signal periods without vehicles involved is given by:

mTnv
Nnv = , (10)
Tr

where Tr is the period of the reference clock signal. On the other hand, with the presence of a vehicle
over the loop, the number of cycles is calculated as:

mTv
Nv = . (11)
Tr

The shift ∆N can then be calculated as the difference between the values given by Equations (10)
and (11) as:
m ∆T
∆N = Nnv − Nv = ( Tnv − Tv ) = m . (12)
Tr Tr
Sensors 2016, 16, 1309 5 of 20

Equating this value to the minimum detection threshold Nt gives us:

∆T Nt Tr
∆N = m = Nt → ∆T = , (13)
Tr m

so from Equation (9), we obtain the threshold sensitivity St as:

∆T Nt Tr Nt
St ∼
=2 =2 =2 . (14)
Tv mTv Nv

From this equation, it can be seen that, for type III detectors, there is a loss in threshold sensitivity
for high oscillation frequencies, although this loss can be easily reduced by increasing the frequency f r
corresponding to the reference clock signal.
Most digital detectors can operate with four or more loops. The problem of crosstalk for resonant
ILDs is solved by separating the loops, in our proposal up to eight directly connected to the detector,
using Time-Division Multiplexing (TDM) [19,21]. The multiplexing could be extended to the detectors
in the surrounding area by using synchronizing signals generated by one of them, which would
work as the master, thus sequencing the time multiplexing of all the detectors. However, this has the
disadvantage of reducing the sampling frequency of the obtained vehicle inductive signatures. These
multiplexed models sequentially feed and analyse the channels more than 100 times per second using
period shift detectors which, as mentioned before, are fast enough to allow these scanning rates.
Our practical implementation is based on a type III detector using TDM, as we will detail in
Section 4. We can see a real vehicle inductive detector of eight loops in Figure 2.

Figure 2. Eight-loop vehicle inductive detector (Afotres–Dimaco).

The oscillation frequency with a negligible resistance R is given by:

1 1
f (t) = = p , (15)
T (t) 2π L(t)C
Sensors 2016, 16, 1309 6 of 20

where the equivalent inductance L(t) is Lnv , with no presence of vehicle, or Lv (t), with vehicle. For a
resonant oscillator, when a vehicle is passing over the loop, both the loop inductance and the oscillation
period decrease.
The signal at the oscillator output is expressed as:

x (t) = Asin (2π f (t)) . (16)

Then, the vehicle inductive signature is the period shift, expressed as ∆T = Tnv − Tv ≥ 0.
Then, the shift in the oscillation period (which gives us the inductive signature) is determined
as follows: p q 
∆T (t) = 2π Lnv C − L(t)C . (17)

2.2. Amplitude ILDs

Amplitude ILDs are based on the measurement of changes in voltage amplitude of an RLC
circuit to which a fixed frequency signal is applied [21,23,24]. The RLC circuit is formed by the loop
inductance L and both the resistance R and the capacity C in the detector, and it is connected to the
sinusoidal voltage generator Vg operating at a fixed frequency f = w/2π. The amplitude of the output
voltage V0 changes with the value of the loop inductance L.
The complex transfer function v(t)/Vg is given by:

v(t) 1
=  , (18)
Vg 1 − jR wL1(t) − wC

and the magnitude of v(t)/Vg , i.e., |v(t)/Vg | is:

|v(t)| 1
= r 2 , (19)
|Vg | 
1
1+ R2 wL(t)
− wC

or equivalently,
2π f L(t)
|v(t)| = |Vg | r  2 . (20)
2 2
R2 1 − (2π f ) L(t)C + (2π f L(t))

This output voltage amplitude could be approximated as a function of the inductance L(t)
as follows:
1
V = |v(t)| ≈ , (21)
k1 L(t) + k2
and the amplitude change is given by:

∆V = Vv − Vnv , (22)

Vv being the voltage amplitude with vehicle presence, and Vnv the voltage amplitude without that
presence. Thus, we have:

1 1 k1 ( Lnv − Lv (t))
∆V = Vv − Vnv = − = ,
k1 Lv (t) + k2 k1 Lnv + k2 (k1 Lv (t) + k2 ) (k1 Lnv + k2 )
∆V k ( Lnv − Lv (t)) k ( Lnv − Lv (t)) /Lnv k1
= 1 = 1 = S. (23)
Vv k1 Lnv + k2 k1 + (k2 /Lnv ) k1 + (k2 /Lnv )
Sensors 2016, 16, 1309 7 of 20

Since Lnv is constant, then the sensitivity S is:

∆L ∼ ∆V
S= =k , (24)
Lnv Vv

i.e., the changes in the inductance of the inductive loop due to the presence of a vehicle modulate
the amplitude of the fixed frequency carrier. In other words, the output voltage signal is
Amplitude Modulated (AM) by the vehicle signature. Therefore, the demodulation of the AM waveform
gives that vehicle signature and also, by means of an Analogue-to-Digital Conversion (ADC), the signature
data. The bandwidth of the vehicle signature is mainly a function of the vehicle speed, the loop
geometry, and the vehicle undercarriage features.
Let n be the bit number for ADC, and therefore N = 2n the state counter. Let also Nt be the count
threshold, and then the threshold sensitivity is expressed as:

Nt
St ∼
=k . (25)
N
The problem of crosstalk for the amplitude ILDs with four or more loops [23] is solved by an
RLC circuit per loop with the carrier frequencies of each loop spaced enough to include the signature
bandwidth, and using a synchronous demodulator tuned to each carrier frequency.

3. Impact of Noise on Digital Detectors

In this section, we present a model to study the impact of noise on both resonant and amplitude
detectors, which have been introduced in Section 2.

3.1. Impact of Noise on Resonant Detectors

Figure 3 shows the block scheme of a resonant detector. Let L(t) be the equivalent inductance on
the ends of the parallel resonant circuit constituted by this inductance and the equivalent capacity C.
The oscillation frequency is given by Equation (15).

x(t) xn(t)
L(t) Resonant Counter
oscillator ∆T(t)

n(t)

Figure 3. Block scheme of a resonant detector.

We will consider interferences caused by Additive White Gaussian Noise (AWGN), denoted as n(t),
induced in the loop by ambient noise, like power lines, emissions from mobile phones, and so on.
Therefore, at the comparator input we have:

x n ( t ) = x ( t ) + n ( t ), (26)

where xn (t) is the signal plus noise. This signal must be converted, previously to be carried out to the
counter input, to a digital pulse train. This conversion is performed by the comparator with hysteresis,
which acts as a wave shaper. Then, the counter receives and counts the m counting cycles and measures
the time interval ∆T, which provides the vehicle inductive signature.
Sensors 2016, 16, 1309 8 of 20

3.2. Impact of Noise on Amplitude Detectors

Figure 4 shows the block scheme of an amplitude detector in presence of noise. Let v(t) be the
signal at the output of the RLC circuit. Again, if we consider interferences produced by AWGN,
denoted as n(t), the signal at the output of the RLC circuit is given by vn (t) as follows:

v n ( t ) = v ( t ) + n ( t ). (27)

The root mean square (rms) value of this signal provides the inductive signature of amplitude ∆A(t).

R Vg

v(t) vn(t) Synchronous Lowpass

∆v(t)
demodulator filter
L(t) ∆T(t)
C
n(t) RMS
∆A(t)

Figure 4. Block scheme of an amplitude detector.

Synchronous Demodulator (SD)

The signal modulated by the inductive signature s(t) can be written as:

x (t) = ( A + s(t)) cos (wt) , (28)

and multiplying x (t) by the carrier, cos(wt), we have:

1 1
y(t) = ( A + s(t)) cos2 (wt) = ( A + s(t)) + ( A + s(t)) cos (2wt) . (29)
2 2
With a low-pass filter, we can eliminate the component of frequency 2w, so that, also removing
the Direct Current (DC) component, the inductive signature ∆v(t) = s(t)/2 is obtained (see Figure 4).

4. Proposed Design of an Inductive Detector

In this section, we present our implementation of the inductive signatures detector, referred
to as SiVIDS. We will describe both hardware and software elements and the procedures for both
measurement and registration of signatures. Our implementation has eight channels, allowing the
registration of signatures of up to four lanes with dual loops in each lane or of up to eight lanes with
simple loops in each lane. This covers most of the existing types of roads and makes the system easy
to build thanks to the availability of a large number of standard integrated circuits with eight channels,
like multiplexers, decoders, buffers, etc.

4.1. Colpitts Oscillator

The oscillation circuit employed in the proposed implementation is the well-known Colpitts
oscillator, since it is the simplest resonant LC oscillator. Figure 5 shows the schematic of a Colpitts
oscillator based on a pnp transistor in common base configuration connected to a tank circuit formed
by the inductance L1 of the inductive loop and the capacitors C1 and C2 that form the capacitive divider
of the feedback loop [25,26].
Sensors 2016, 16, 1309 9 of 20

The oscillation frequency is determined by the parallel resonant circuit formed by the inductance
L1 , and the equivalent capacitor C obtained from the serial connection of C1 and C2 , i.e.,

1 C1 C2
f0 = √ , with C = . (30)
2π L1 C C1 + C2

V1=+5V

R3 R2 C2
R1
L1
Q1 C1

Vo

Figure 5. Circuit for the Colpitts oscillator.

4.2. Pulse Counter

Figure 6 shows a block diagram of the comparison and capture process necessary to measure
the oscillation period automatically. The pulses from the oscillation loop that has been selected as
input are carried to a counter input, so that when a fixed number of pulses m is reached, the measured
value N is captured from a timer working at the frequency f r of the reference clock signal.

Number of counting
pulses

Comparator Match
control
Pulses from
oscillators Pulse counter
Interrupt
Reference Timer in capture
clock mode

Measured value

Figure 6. Measurement method by using comparison and capture.

Since the basic measurement process is performed by hardware using interruptions, the delay
time of interruption attention (known as latency) is not critical.
Figure 7 shows the practical implementation of the multiplex system with eight coils. It consists
of eight Colpitts oscillators connected to eight inductive loops and an analogue multiplexer, which
selects, at each instant, the oscillation signal of one of the loops using a decoder circuit. The use of
eight oscillators instead of a single one allows us to avoid the introduction of an analogue multiplexer
into the oscillation loop, which would be an additional error source.
The output signal of the multiplexer is carried to a shaping circuit, which converts the sinusoidal
signal at its input into a digital pulse. That digital pulse is the input at the counter in the micro
controller, which manages the entire system.
Sensors 2016, 16, 1309 10 of 20

Due to the large amount of data captured by the system, a Compact Flash (CF) memory is employed
for the recording of the signatures that will be subsequently analysed by a computer using signal
processing algorithms. This off-line processing will allow us to perform vehicle classification and
measurement of parameters such as speed or length, and even vehicle re-identification for monitoring
and control applications of vehicular traffic.
For the implementation of our system, we have chosen the AT89C51RE2 micro controller
(Atmel, San Jose, CA, USA) since, firstly, it incorporates the comparison and capture unit needed
in our application; and secondly, it can be easily interconnected to a CF memory bus. Figure 8
shows the interconnections for the AT89C51RE2 micro controller in our system. The output of the
multiplexer is connected to the EC1 input, the CEX0 comparison output is carried to the T2EX input
for capturing/interrupting, and the T2 timer is in capture mode.

SL1

Loop 1 Oscillator 1 SL1

SL2
SL2 SL3
SL4 1-of-8
Loop 2 Oscillator 2 SL5 Decoder
SL6
SL7
SL3
SL8 ABC
Loop 3 Oscillator 3

SL4
A B C
Loop 4 Oscillator 4 Micro
8 Channels controller
analog
SL5 multiplexer Counter
+
Loop 5 Oscillator 5 Shaper
Timer
SL6

Loop 6 Oscillator 6 fr= 4MHz

SL7
CF
Loop 7 Oscillator 7 memory

SL8

Loop 8 Oscillator 8

Figure 7. Multiplex system with eight inductive loops.

AT89C51RE2

Oscillator
ECI
loop pulses

CEX0
Interrupt T2

Match T2EX
control

Figure 8. Interconnections for AT89C51RE2.

Sensors 2016, 16, 1309 11 of 20

Figure 9 shows a picture of the implemented hardware prototype. The left side of the board
includes the eight oscillation circuits with multiplexing and a 16-pin connector for the connection
of the eight inductive coils. The right side contains the micro controller and a Real-Time Clock (RTC)
circuit with a lithium battery providing the date and time. The CF memory card used for the storage of
the captured signatures can be seen at the bottom. One of the main advantages of the proposed system
is that it can be implemented at a very low cost, thanks to its simplicity.

Figure 9. Photo of the hardware prototype.

4.3. Measurement
As can be seen in Equation (12), the period of the oscillation signal in each loop is calculated as
the number of cycles N of the reference clock signal in m cycles of the oscillation signal of that loop.
The measurement of N is made by means of T2 interruptions generated by overflow (TF2) and by
hardware automatic capture (T2EX). An initial number of oscillation cycles mi corresponding to the
stability time of the oscillator start are discarded.
Figure 10 shows a flowchart describing the process of the T2 interruptions’ attention.
The measurement of each loop starts with the interruption by T2 overflow due to the delay time
between loops required for the oscillation of the previous loop to completely disappear (the branch
with number 1 in the figure). At that point, a new measurement loop is selected, the mi value is
initialized to the number of initial start cycles, the maximum time for the measurement is established,
and the corresponding oscillator is started.
Next, the branch marked with 2 in the figure is executed, so that the time Ni at initial start cycles
is measured and the number m of counting cycles to be measured is loaded.
Finally, the measurement process finishes with the interruption by T2 capture when m is reached,
which corresponds to branch 3 in Figure 10. In this moment, the time interval between loops is loaded,
and the oscillation loop stops and saves the measured time N obtained after subtracting the value Ni
of step 2.
Branch 4 only occurs in the unlikely event that the loop has problems with the start of the
oscillation. In such a case, the inter-loop waiting time is loaded and N is set to zero, which indicates
that the loop is not oscillating.
Sensors 2016, 16, 1309 12 of 20

Start of T2
interruption attention

No Yes
T2 Capture T2 Overflow? T2 Overflow

No Yes No Yes
Last Counting? Last Counting?
2 3 4 1

Save start Load waiting time Load waiting time Increase number of
counting time for ready loop for ready loop measured loop

Load comparison Stop oscillation Save measured Load comparison

value of measured loop time value

Save measured Load maximum

time measurement time

Run oscillator of
measured loop

End of T2
interruption attention

Figure 10. Flowchart of attention at T2 interruption.

4.4. Registration
The oscillation period of the coils is continuously measured to determine the reference value of
each coil at rest, i.e., without the presence of a vehicle. With the goal of adapting to the variations
in the environmental conditions suffered by the coils, an adaptive algorithm, similar to the one
described in [20], is employed. This algorithm tries to correct the reference value according to such
external factors.
When the measured period of a coil is less than its reference value, which means that a vehicle is
over the coil, the corresponding entry is made in the internal memory, storing the inductive signature
of the vehicle.

Figure 11. A photo of the measurement location, with GPS coordinates: 43.235941 (Lat.);
−8.464462 (Long.).
Sensors 2016, 16, 1309 13 of 20

In order to test the hardware prototype we captured vehicle inductive signatures in Río Anllóns
station in the AC-523 road (Ledoño-Meirama, Spain), kilometre 7. A picture of this location is shown in
Figure 11. The detector equipment was located inside the cabinet of the Río Anllóns station, also shown
in the photo. Since the road is two lanes wide (one for each direction), we have connected four inductive
loop sensors, two on each side of the road. These sensors are squares with a side length of 2 m and a
distance between their centres of 5 m (Figure 12).

Lane 2 Loop4 Loop3

Lane 1 Loop1 Loop2

Detector

Figure 12. Interconnections from the inductive loop to the detector.

5. Experimental Section
The inductive loop sensors work at a sampling period of T = 10 ms, or, equivalently, a sampling
frequency of 100 Hz. In any case, the sampling frequency could be easily increased by increasing the
frequency of the reference clock, if required by other measurement conditions. Moreover, waiting times
between coils are needed to guarantee that the oscillation vanishes in a coil before starting in the
next one.
Figure 13 illustrates the measured oscillation in a coil operating at a frequency of 56.2 kHz,
using mi = 10 initial cycles (in red) and m = 35 measurement cycles (in green). The aforementioned
waiting times at the beginning and at the end of the oscillations can also be observed (in blue) in
the figure.
4

1
Amplitude (V)

-1

-2

-3

-4
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (ms)

Figure 13. Capturing of a measurement of the coil oscillation.

With the described sensors, two inductive signatures are obtained from each passing vehicle.
These inductive signatures will be very similar, although there can be small differences due to the fact
Sensors 2016, 16, 1309 14 of 20

that they are taken in different loops and time instants, and, in general, also for different positions and
accelerations of the vehicle. In order to display those signatures, we have developed a software tool
using the development environment for a visual programming language Labview [27,28]. This tool
allows us to download the file containing the signatures from the CF card. In addition to the ILDs,
we placed a video camera for the recording of the passing vehicles, so we could associate each vehicle
to its corresponding inductive signature.
It is important to note that we have acquired the real inductive signatures ∆T (t) using our SiVIDS
prototype, and then from Equation (17), the equivalent inductance L(t) of each signature has been
obtained. For the evaluation of the impact of the noise on resonant and amplitude detectors, AWGN
noise will be added to the signals x (t) of Equation (16) and v(t) of Equation (18), respectively.
In our implementation, the Colpitts oscillator explained in Section 4.1 has C1 = C2 = 100 nF
and therefore C = 50 nF. Thus, an inductance of L1 = 100 µH results in an oscillation frequency of
71.18 kHz. The circuit has been designed for oscillation frequencies within the range [25 kHz, 100 kHz]
i.e., for coils with inductances between 50 µH and 800 µH. Therefore, it is not necessary to adjust the
frequency of the LC oscillators (known as tuning).

6. Results and Discussion

In this section, we will show some results obtained from the inductive signatures captured using
the prototype presented in this paper.

Figure 14. From left to right, the upper figures show the photo of a car and the respective signatures
obtained with the Loops 3 and 4. The figures in the middle of the picture show a truck and its
corresponding signatures captured also using the Loops 3 and 4. The lower figures display the photo
of a bus and two signatures obtained with the first and the second loop, respectively.
Sensors 2016, 16, 1309 15 of 20

Figure 14 shows the photos of three different vehicles and their corresponding inductive signatures
obtained in the dual loops, as an example of the more than one thousand inductive signatures captured
with our system. As it can be seen in the figure, there is a great similarity between the pair of signatures
of any of the vehicles, in contrast with the significant difference in the signatures obtained for different
types of vehicles. Thus, each type of vehicle (car, truck, bus...) can be classified under a unique
inductive signature, which will depend on the parameters that define each of them, such as size,
distribution of the metal mass, engine and axle location, spacing between the undercarriages and the
road, etc.

6.1. Effect of Noise on Vehicle Inductive Signatures

Figure 15 shows the real vehicle inductive signature obtained with noise—for a Signal-to-Noise
Ratio (SNR) of 15 dB—and with no presence of noise in the system. As we can see in the figure, even
for 15 dB of SNR, the shape of the noisy inductive signature is quite similar to that obtained with no
noise at the detector input, which verifies that our resonant detector is robust against environmental
noise. Figures 16 and 17 show the impact of noise on the amplitude detectors for the same SNR.

25
SNR 15 dB
No noise

20

15
∆T(µs)

10

0
0 5 10 15 20 25 30 35
Samples

Figure 15. Inductive signature of the resonant detector with and without noise.

0.3
SNR15 dB
No noise
0.25

0.2
∆A(V) rms

0.15

0.1

0.05

-0.05
0 5 10 15 20 25 30 35
Samples

Figure 16. Inductive signature of the root mean square (rms) amplitude detector with and without noise.
Sensors 2016, 16, 1309 16 of 20

0.9
SNR 15 dB
0.8 No noise

0.7

0.6

0.5

∆V(V) 0.4

0.3

0.2

0.1

-0.1
0 5 10 15 20 25 30 35
Samples

Figure 17. Inductive signature of the detector with synchronous demodulator with and without noise.

Finally, we compare the performances of both resonant and amplitude detectors, in terms of SNR at
the detector output. For this purpose, we calculate the output SNR as follows: firstly, we determine the
level of signature signal without noise, i.e., ∑ ∆x f ; then, the level of noise is obtained as ∑ |∆x − ∆x f |,
where ∆x is the signature signal plus noise; Finally, the output SNR is calculated as:

∑ ∆x f
output SNR = 20log . (31)
∑ |∆x − ∆x f |

The average output SNR is obtained by considering 556 real inductive signatures captured in the
AC-523 road with the resonant detector, so that then the equivalent inductance L(t) of each signature
is obtained. Figure 18 shows the average output SNR as a function of the input SNR for resonant and
amplitude detectors. This figure shows a good behaviour against noise of the resonant detector for
input SNR greater than 12 dB, even better than that obtained for the amplitude detector. However, for
SNRs lower than 12 dB, the resonant detector is very sensitive to noise and, therefore, not useful for
the purposes described in this work.

50
Resonant
40 RMS Amplitude
AM Demodulator
30

20
Output SNR (dB)

10

-10

-20

-30

-40
0 5 10 20 25 30
Input SNR (dB)

Figure 18. Average output Signal-to-Noise Ratio (SNR) for resonant and amplitude detectors.
Sensors 2016, 16, 1309 17 of 20

6.2. Effect of Noise on Speed Estimation and Vehicle Classification

In this subsection, we will evaluate the effect of the noise on two different applications usually
required for ITS: speed estimation and vehicle classification.
According to Figure 19, the following time instants using double loop are determined:

• t1 : input time instant of the normalized vehicle signature 1;

• t2 : output time instant of the normalized vehicle signature 1;
• t3 : input time instant of the normalized vehicle signature 2;
• t4 : output time instant of the normalized vehicle signature 2.

Loop 1 Loop 2

Signature 1 Signature 2
1

t
t1 t2 t3 t4

Figure 19. Scheme for time measurements.

The standard method for speed estimation uses the following expression, according to the
aforementioned notation [29,30],
 
1 d d
ŝ = + , (32)
2 t3 − t1 t4 − t2

where d is the distance between loop centres. From this expression, the vehicle length can be directly
obtained using the following estimator:

( t2 − t1 ) + ( t4 − t3 )
L̂ = ŝ × − w, (33)
2
with square loops of side length w. For vehicle classification, the vehicles passing on the road will be
classified using a threshold-based criterion as indicated in Table 1.

Table 1. Vehicle classifiation using a length-based criterion, i.e., based on the L̂ estimator of Equation (33).

Vehicle Classification Small Medium Large

Type of vehicles Car Large car, van Truck, bus, trailer
Number of vehicles (manually pre-classified with a video camera) 680 61 168
Decision rule L̂ ≤ e1 e1 < L̂ ≤ e2 L̂ > e2

Notice that e1 and e2 are the thresholds empirically obtained from a training stage. For this
training, only the Loops 1 and 2 are used. The value corresponding to the threshold e1 was obtained
Sensors 2016, 16, 1309 18 of 20

by maximizing the success rate in vehicle classification when only small and medium vehicles are
considered, while the threshold value e2 is the result of a similar maximization when only medium and
large vehicles are computed. The optimum values for those thresholds are e1 = 5.6 m and e2 = 6.5 m.
Applying these methods, Figure 20 shows the influence of AWGN on vehicle classification
with the length-based criterion and on vehicle speed estimation. The error percentage for vehicle
classification is calculated as follows:
cAWGN
errorc (%) = 100 × , (34)
c0

where cAWGN is the total number of misclassified vehicles under the presence of AWGN, calculated with
respect to the classification without noise, and c0 is the total number of vehicles. On the other hand,
the percentage of error for speed estimation is calculated as follows:

|sAWGN − s0 |
errors (%) = 100 × , (35)
s0

where sAWGN and s0 are, respectively, the estimated vehicle speeds with and without the presence of
AWGN. All the results have been averaged for the dataset collected from the AC-523 road.
This figure allows us to conclude that in both applications the effect of noise is almost negligible
for SNR above 12 dB.
150
Classification
Speed

100
Average error (%)

50

0
4 6 8 10 12 14 16 18 20
Input SNR (dB)

Figure 20. Influence of Additive White Gaussian Noise (AWGN) on vehicle classification and speed
estimation errors.

7. Conclusions
Experimental results have shown that ILD sensitivity is extremely repeatable for fixed sizes and
geometries of both the loop and the vehicle, and for a fixed distance between them. Therefore, we have
shown that the sensitivity can be approximated by variations of period and amplitude of voltage of
the oscillation signal. Moreover, resonant ILDs show an adequate compromise between reliability and
cost, which determines that such detectors have been selected for our practical implementation.
In this paper, we have presented a simple module for the capture of inductive vehicle signatures
based on TDM. The implemented system performs a sequential scanning using analogue multiplexing
of up to eight oscillators and detects the presence of a vehicle by means of a shift in the period of the
signals from the selected oscillator. It subsequently captures the inductive signature of the detected
vehicle by measuring the time it needs to count a fixed number of pulses.
Sensors 2016, 16, 1309 19 of 20

In the experimental results obtained from measurements in a real scenario using dual loops,
we observed a good similarity between the pair of signatures obtained from the same vehicle and
a significant difference between the signatures corresponding to different vehicles, which validates
the good performance of our implementation and enables its use in applications such as vehicle
classification, speed and length measurement using only one loop, and re-identification of vehicles for
supervision and control tasks in vehicular traffic.
Moreover, the performance of the resonant detector proposed in this work is validated in the
presence of AWGN determining an input SNR higher than 12 dB.

Acknowledgments: The authors wish to thank Xunta de Galicia and the Ministerio de Economía y Competitividad
of Spain for their financial support with grants R2014/037 and TEC2013-47141-C4-1-R, respectively. The authors
also thanks Dimaco S.C.G. (A Coruña) for the infrastructures on which we carried out the experimental tests in
real scenarios.
Author Contributions: J.J. Lamas-Seco and A. Dapena conceived and designed the experiments; J.J. Lamas-Seco
and F.J. Vazquez-Araujo performed the experiments; P.M. Castro and F.J. Vazquez-Araujo analysed the data;
P.M. Castro and A. Dapena wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Klein, L.; Gibson, D.; Mills, M. Traffic Detector Handbook, 3rd ed.; FHWA-HRT-06-108; Federal Highway
Administration, Turner-Fairbank Highway Research Center: McLean, VA, USA, 2006; Volume 1.
2. FHWA. Measurement of Highway-Related Noise. Available online: http://www.fhwa.dot.gov/
environment/noise/measurement/mhrn03.cfm (accessed on 6 July 2011).
3. FHWA. A Summary of Vehicle Detection and Surveillance Technologies use in Intelligent Transportation
Systems. Available online: https://www.fhwa.dot.gov/policyinformation/pubs/vdstits2007/05.cfm
(accessed on 7 November 2014).
4. Gajda, J.; Sroka, R.; Stencel, M.; Wajda, A.; Zeglen, T. A vehicle classification based on inductive loop
detectors. In Proceedings of the 18th IEEE Instrumentation and Measurement Technology Conference,
Budapest, Hungary, 21–23 May 2001; Volume 1, pp. 460–464.
5. Jeng, S.T.; Chu, L.; Hernandez, S. Wavelet-k Nearest Neighbor Vehicle Classification Approach with
Inductive Loop Signatures. Transp. Res. Rec. J. Transp. Res. Board 2013, 2380, 72–80.
6. Jeng, S.T.; Ritchie, S. Real-Time Vehicle Classification Using Inductive Loop Signature Data. Transp. Res. Rec.
J. Transp. Res. Board 2008, 2086, 8–22.
7. Ki, Y.; Bai, D. Vehicle Classification Model for Loop Detectors Using Neural Networks. Transp. Res. Rec. J.
Transp. Res. Board 2005, 1917, 164–172.
8. Zhang, G.; Wang, Y.; Wei, H. Artificial Neural Network Method for Length-Based Vehicle Classification
Using Single-Loop Outputs. Transp. Res. Rec. J. Transp. Res. Board 2006, 1945, 100–108.
9. Ndoye, M.; Totten, V.; Carter, B.; Bullock, D.; Krogmeier, J. Vehicle Detector Signature Processing and Vehicle
Reidentification for Travel Time Estimation. In Proceedings of the Transportation Research Board 87th
Annual Meeting, Washington, DC, USA, 13–17 January 2008.
10. Oh, C.; Tok, A.; Ritchie, S. Real-time freeway level of service using inductive-signature-based vehicle
reidentification system. IEEE Trans. Intell. Transp. Syst. 2005, 6, 138–146.
11. Sun, C.; Ritchie, S.G.; Tsai, K.; Jayakrishnan, R. Use of vehicle signature analysis and lexicographic
optimization for vehicle re-identification on freeways. Transp. Res. 1999, 7C, 167–185.
12. Tawfik, A.Y.; Abdulhai, B.; Peng, A.; Tabib, S.M. Using Decision Trees to Improve the Accuracy of Vehicle
Signature Reidentification. Transp. Res. Rec. J. Transp. Res. Board 2004, 1886, 24–33.
13. Wang, Y.; Nihan, N.L. Freeway Traffic Speed Estimation with Single-Loop Outputs. Transp. Res. Rec. J.
Transp. Res. Board 2000, 1727, 120–126.
14. Sun, C.; Ritchie, S. Individual Vehicle Speed Estimation Using Single Loop Inductive Waveforms.
J. Transp. Eng. 1999, 125, 531–538.
15. Mills, M. Inductive loop detector analysis. In Proceedings of the 31st IEEE Vehicular Technology Conference,
Washington, DC, USA, 6–8 April 1981; Volume 31, pp. 401–411.
 Sensors 2016, 16, 1309 20 of 20

16. Mills, M. Inductive loop system equivalent circuit model. In Proceedings of the IEEE 39th Vehicular
Technology Conference, San Francisco, CA, USA, 1–3 May 1989; Volume 2, pp. 689–700.
17. Cheevarunothai, P.; Wang, Y.; Nihan, N. Identification and Correction of Dual-Loop Sensitivity Problems.
Transp. Res. Rec. J. Transp. Res. Board 2006, 1945, 73–81.
18. Day, C.; Brennan, T.; Harding, M.; Premachandra, H.; Jacobs, A.; Bullock, D.; Krogmeier, J.; Sturdevant, J.
Three-Dimensional Mapping of Inductive Loop Detector Sensitivity with Field Measurement. Transp. Res.
Rec. J. Transp. Res. Board 2009, 2128, 35–47.
19. Martin, M. Microprocessor Controlled Loop Detector System. U.S. Patent 4680717, 14 July 1987. Available
online: http://www.google.co.in/patents/US4680717 (accessed on 14 July 1987).
20. Minsen, C.; Ngarmnil, J.; Rongviriyapanich, T. Embedded adaptive algorithm for multi-lanes-traffic
inductive loop detecting system. In Proceedings of the 2005 2nd International Conference on Electrical
Engineering/Electronics, Computer, Telecommunications and Information Technology, Pattaya, Thailand,
12–13 May 2005; Volume 1, pp. 359–362.
21. Sheik Mohammed Ali, S.; George, B.; Vanajakshi, L.; Venkatraman, J. A Multiple Inductive Loop Vehicle
Detection System for Heterogeneous and Lane-Less Traffic. IEEE Trans. Instrum. Meas. 2012, 61, 1353–1360.
22. Gajda, J.; Stencel, M. A Highly Selective Vehicle Classification Utilizing Dual-Loop Inductive Detector.
Metrol. Meas. Syst. 2014, 21, 473–484.
23. Hilliard, S.; Roberts, M.; Yerem, G. Inductive Signature Measurement System. U.S. Patent 6911829, 28 June
2005. Available online: http://www.google.tl/patents/US6911829 (accessed on 28 June 2005).
24. Sheik Mohammed Ali, S.; George, B.; Vanajakshi, L. An Efficient Multiple-Loop Sensor Configuration
Applicable for Undisciplined Traffic. IEEE Trans. Intell. Transp. Syst. 2013, 14, 1151–1161.
25. Malik, N.; García, M.; Ordas, M.; Viejo, C. Circuitos Electrónicos: Análisis, Diseño y Simulación;
Pearson Educación: Mexico City, Mexico, 1996.
26. Blake, R. Sistemas Electrónicos de Comunicaciones; International Thomson: Boston, MA, USA, 2004.
27. Travis, J.; Kring, J. LabVIEW for Everyone: Graphical Programming Made Easy and Fun; Prentice Hall:
Englewood Cliffs, NJ, USA, 2006.
28. Blume, P.A. The LabVIEW Style Book; Prentice Hall: Englewood Cliffs, NJ, USA, 2007.
29. Ki, Y.K.; Baik, D.K. Model for accurate speed measurement using double-loop detectors. IEEE Trans.
Veh. Technol. 2006, 55, 1094–1101.
30. Gordon, R.L.; Tighe, W. Traffic Control Systems Handbook; Report No.: FHWA-HOP-06-006; Federal Highway
Administration: Washington, DC, USA, 2005.

c 2016 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/).

View publication stats