Photonics-Based Dual-Band Radar for Landslides
Monitoring in Presence of Multiple Scatterers
Suzanne Melo , Salvatore Maresca, Sergio Pinna , Filippo Scotti , Milad Khosravanian,
Arismar Cerqueira S. Jr.
, Filippo Giannetti , Abhirup Das Barman, and Antonella Bogoni
Abstract—In this paper, a dual-band photonics-based radar
system used for precise displacement measures in a multitarget sce-
nario is described. The radar was designed for monitoring applica-
tions to prevent both structural failures of buildings and landslides.
The radar system exploits the technique of stepped frequency con-
tinuous wave signal modulation and the displacement of the targets
is evaluated through differential phase measurements. In this work,
encouraged by the results already achieved in the single-target sce-
nario, we present an investigation extended to the case of multiple
targets. We aim to evaluate the accuracy of the displacement esti-
mation both from a simulated and experimental point of view, and
to understand how multiple targets impact on the final estimate of
displacements. Simulation results demonstrate that it is possible to
achieve a typical accuracy of less than 0.2 mm for distances up to
400 m. These results are confirmed by preliminary experimental
outcomes, which take into account different operative conditions
with multiple targets. Finally, concluding remarks and perspectives
draw the agenda for our future investigations.
Index Terms—Interferometry, mode-locked lasers, radio
frequency photonics, radar, remote sensing and sensors.
I. INTRODUCTION
A
LARGE branch of civil engineering is interested in build-
ing integrity analysis and slope stability monitoring appli-
cations. For these, it is of crucial importance to be able to detect
tiny target displacements which could be precious indications
of forthcoming structural failures or landslides [1].
This task has been accomplished in many different ways over
the years. For example, using extensometers, inclinometers,
Manuscript received August 30, 2017; revised January 13, 2018; accepted
March 2, 2018. Date of publication March 9, 2018; date of current version
March 23, 2018. This work was supported in part by the National Project PRE-
VENTION (with the contribution of Ministry of Foreign Affairs, Directorate
General for the Country Promotion) and in part by the EU projects ROBORDER
#740593. (Corresponding author: Suzanne Melo.)
S. Melo, S. Maresca, and A. Bogoni are with the Scuola Superiore
Sant’Anna, 56124 Pisa, Italy (e-mail:, s.assisdesouzamelo@sssup.it; salvatore.
maresca@cnit.it; antonella.bogoni@sssup.it).
S. Pinna is with the Photonics Laboratory, University of California, Santa
Barbara, CA 93106 USA (e-mail:,pinna@ece.ucsb.edu).
F. Scotti is with the CNIT (Consorzio Nazionale Interuniversitario per le
Telecomunicazioni), 56124 Pisa, Italy (e-mail:,filippo.scotti@cnit.it).
M. Khosravanian and F. Giannetti are with the Department of Infor-
mation Engineering, University of Pisa, 56122 Pisa, Italy (e-mail:, soroor
_2069@yahoo.com; filippo.giannetti@iet.unipi.it).
A. Cerqueira S. Jr. is with the National Institute of Telecommunications
(Inatel), Santa Rita do Sapucai 37540-000, Brazil (e-mail:,arismar@inatel.br).
A. Das Barman is with the Institute of Radio Physics and Electronics, Uni-
versity of Calcutta, Kolkata 700 009, India (e-mail:,abhirup1.rpe@gmail.com).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JLT.2018.2814638
crack meters, Global Positioning System (GPS) receivers, etc.
However, their ability to provide a deformation map over large
areas is restricted, due to their localized operation. In this con-
text, monitoring based on radars can be quite advantageous.
Radars can largely overcome previous traditional methods in
terms of remotely operation, in case of difficult and inaccessible
areas, and the ability to offer deformation maps of large areas.
In this context, radar differential interferometry is one of the
most effective ways for measuring such structural movements,
providing displacement maps with very high spatial resolution,
especially for disaster prevention and risk mitigation [1], [2].
The majority of radar systems deployed for such applica-
tions are made up by airborne or spaceborne synthetic aperture
radar (SAR) and by ground-based systems. SAR systems ex-
ploit the motion of the radar antenna to provide finer spatial
resolution w.r.t. conventional beam-scanning radars. Typically,
they are mounted on moving platforms (e.g., aircraft or satel-
lite). Thus, the high spatial resolution obtained through SAR
processing and the wide coverage area offered by the satellite
platform allow to obtain high resolution displacement maps of
wide land areas [3]. However, these features come at the cost
of long revisiting time (no real-time monitoring on a continu-
ous basis), unwrapping problems (i.e., false target problems),
and signal phase distortions caused by the propagation in the
atmosphere [4]. Conversely, ground-based systems are precious
tools for monitoring smaller areas without interruption and can
provide precise displacement estimates down to sub-millimeter
scale at hundreds of meters up to few kilometers distance [5].
A common approach for estimating target displacements
is to use differential interferometry, exploiting the Stepped-
Frequency Continuous Wave (SFCW) signal modulation [5].
Differently from the pulsed radar where a pulse is transmit-
ted and time-of-flight measurements are used to determine the
distance from objects, CW radars continuously broadcast radar
waveforms, which may be considered to be pure sine waves, and
the phase of the returning echo is used to determine range. In
SFCW radars a sequence of frequency steps may be employed
to extend the range. Displacement evaluation are obtained by
interferometric technique, i.e., comparing the phase information
of the back scattered signal collected at different times.
Initially developed to detect buried objects, SFCW radar sen-
sors are, nowadays, widely used in a number of both civilian
and military applications. In particular, detection and identifi-
cation of dangerous conditions of civil strictures (e.g., bridges,
buildings, buried pipes or archeological artifacts).
In SFCW radars, a sequence in time domain of sinusoidal sig-
nals with slightly different frequencies is generated and trans-
mitted for a fixed time, named step interval generating pulsed
trains. These sinusoidal signals are coherently generated with a
constant frequency separation, namely the frequency step, in a
defined frequency range [6]. The total used band is given by the
difference between the the minimum and maximum frequencies
of sinusoids and the frequency step width are jointly chosen to
satisfy a given operative unambiguous range and range resolu-
tion, as better described in Section III. Several cycles of step
frequency modulation are used for each acquisition. Each si-
nusoid composing the SFCW signal is characterized by a very
narrow instantaneous bandwidth. For this reason, this solution
results in a valuable improvement of the noise figure at the re-
ceiver, thus leading to an improvement of both sensitivity and
dynamic range, whilst maintaining a good average power. More-
over, by transmitting only one frequency at a time, it could be
possible to properly compensate, through ad-hoc signal process-
ing techniques, the received signals, in the case they propagate
through lossy media with known characteristics, and for any
non-linear effect caused by the transmitter or the receiver. Fur-
thermore, one of the main advantages of SFCW modulation is
that the signal energy is distributed in time. This feature al-
lows to reduce the system power requirement, and it is ideal
to avoid non-linear effects of the electronic components due to
high power transmission [7].
Once the backscattered signal is acquired by the receiver, it
is mixed with the reference signal (i.e., the transmitted signal),
in order to obtain the phase information. Then, differential in-
terferometry is used to evaluate the phase variations between
two consecutive target observations. Phase information is then
converted into a precise displacement measure, as will be better
explained in Section III. It is straightforward that the system
phase stability and the coherence among the frequency tones
play a decisive role in the final accuracy of phase measure-
ments. Therefore, special attention should be aimed to limit the
phase noise in the signal generation stage, thus guaranteeing a
high degree of signal coherence.
An excellent solution is based on the use of photonics. In fact,
recently photonics for microwave systems have been demon-
strated to guarantee a higher stability than electronic technology.
In fact, photonics enable the generation and reception of Radio
Frequency (RF) signals, without the need of noisy mixers and
RF oscillators. Mixers are replaced by low-noise electro-optical
and opto-electronic conversion blocks, while RF oscillators are
substituted by highly stable optical oscillators (i.e., lasers), ac-
cordingly improving the RF signal coherence [8]. In addition,
photonics help in improving the frequency flexibility of the
system. In fact, with a single photonics-based transceiver (i.e., a
single optical oscillator) it is possible to simultaneously generate
multiple signals in different RF bands, as demonstrated in [9].
This is a very desirable feature, because it guarantees an intrinsic
phase coherence among the generated multiband signals.
In this framework, our group proposed for the first time a
photonics-based radar system able to exploit the combination
of SFCW modulation and differential interferometry to perform
high precision displacement measurements [10]. At its actual
stage of development, the prototype is a dual-band radar, mean-
ing that it can contemporaneously generate and transmit two
signals, coherent with each other, and locked on two different
frequency bands (S- and X-bands).
Coherent multi-band radars introduce important innovative
and advantageous features. In fact, the possibility to choose dif-
ferent carrier frequencies makes the system tunable, and, thus,
more sensitive to the different scattering properties of a given
target. Moreover, the choice of multiple frequencies makes the
system robust against unwanted returns from the environment.
These returns, called clutter, can respond in different ways ac-
cording to several factors (e.g., the size of the scatterer w.r.t. the
wavelength, the geometry, etc.), generating false alarms (i.e.,
false targets). Weather conditions can have an impact on the re-
ceived signal as well, in terms of echo attenuation or generation
of false targets. At last, as known from the radar range equation,
the carrier frequency has an effect on the the final reachable
range of interest, due to the different absorptions and distor-
tions experienced by the RF signal as a f unction of its carrier
frequency [11].
Another innovative and useful feature of the photonics-based
solution is the strong coherence among the generated sinusoidal
tones of the SFCW signal, which allows to minimize the phase
fluctuations between two consecutive observations, i.e., radar
measurements. Consequently, an improvement of the displace-
ment measure accuracy is obtained without using complex cor-
rection techniques. With low phase noise, the phase variations
between two consecutive measurements will be, almost entirely,
proportional to the range shift we want to estimate by means
of the interferometric algorithm. Precision better than 200 μm
up to 3 km distance was obtained [10]. This achievement is
strongly welcome for a timely and reliable early detection of
structural risks. Moreover, the possibility to replicate the SFCW
modulation in several bands (two in our case) allows to highly
increase the maximum spectral spacing of the tones, improving
the displacement resolution obtained with the interferometric
processing [10].
Here, we push forward the study proposed in [10] and ex-
tend the one presented in [12], and we investigate the system
capabilities in a real multiple target scenario, with the aim to
understand how the presence of multiple targets can affect the
precision of the single displacement estimates. In this work, we
present and discuss both numerical and preliminary experimen-
tal results. The obtained results show a good agreement, as it
will be discussed in the following of the paper. The outline is as
follows: section II introduces the photonics-based radar system,
while Section III deals with the theory of differential phase es-
timation. Sections IV and V present and discuss the numerical
simulations and the experimental results, respectively. Finally,
Section VI summarizes the main outcomes of the proposed work
and recapitulates the results we achieved. Further guidelines are
provided as well, which will drive our future research.
II. T
HE PHOTONICS-BASED RADAR SYSTEM
The concept of the dual-band photonics-based radar for build-
ing integrity analysis and slope stability monitoring applications
Fig. 1. Concept of the photonics-based dual-band radar system for the monitoring of mountain slope stability and building integrity. In the shaded blue box are
included the photonics-based subsystems.
is illustrated in Fig. 1. The radar illuminates the scene transmit-
ting contemporaneously on two separate carrier frequencies,
precisely in the S- (2.475 GHz) and X-bands (9.875 GHz).
The two transmitted RF signals are generated at intermediate
frequencies (IFs) by a digital waveform generator. The IF sig-
nals are then fed to the photonics-based transceiver to be up-
converted at the RF carrier frequencies. The RF signals, upon
amplification by a wideband RF amplifier (WBA), are transmit-
ted towards the targets through a wideband horn antenna. The
echoes, collected by a second receiving horn antenna, are first
down-converted to IFs by the photonics-based transceiver and
then processed by a digital signal processing (DSP) unit.
In our setup, each target lays in a different range cell and can
be independently moved through a high precision digitally con-
trolled motorized linear platform. The motorized slide, remotely
controlled using a laptop, allows a precise target positioning.
The principle of the photonics-based transceiver is depicted
in the shaded blue box in Fig. 1, and detailed in the shaded blue
box in Fig. 2 (left). Instead of using RF clocks, a single optical
clock is exploited, represented by a mode-locked laser ( MLL).
The MLL is a highly stable pulsed laser, and its spectrum, as
shown in Fig. 2(a), is composed by series of modes spaced by
the laser pulse repetition frequency F
MLL
= 400 MHz. For the
frequency up-conversion process, the two signals to be trans-
mitted in S- and X-bands, are generated at IF
S
= 75 MHz and
IF
X
= 125 MHz respectively, as depicted in Fig. 2(b). Then,
they are transferred into the optical domain modulating the MLL
optical pulses by means of a Mach-Zehnder modulator (MZM).
In fact, using the IF signals as modulating signal, their replicas
are generated in the optical domain as lower and upper side-
bands of each spectral line of the MLL at IF distance, as shown
in Fig. 2(c).
To avoid aliasing, the sum of the IF
S
and IF
X
frequencies
and their respective spectral occupancy must be chosen
less or equal than F
MLL
/2. At the output of the MZM, the
heterodyning of all the optical spectral components into a
photodiode (PD) produces a replica of the two IF signals at
every CF
S
= k · F
MLL
± IF
S
and CF
X
= l · F
MLL
± IF
X
frequency, respectively, where k and l are non-negative integers,
see Fig. 2(d). At this point, at the output of the PD it is possible
to contemporaneously transmit the required up-converted
signals at CF
S
= 2.475 GHz and CF
X
= 9.875 GHz, by using
two separate electrical band-pass filters (BPF) locked at the
two desired S- and X-band carrier frequencies, as illustrated in
Fig. 2(e). Moreover, it is important to observe that the generated
carrier frequencies are limited only by the PD bandwidth.
Afterwards, the received RF signal is transferred again to the
optical domain, through a second MZM, which modulates the
same optical pulse train generated by the MLL. This operation
generates replicas of the received RF waveforms around each of
the MLL modes. The final spectrum we obtain is, thus, analog to
the one represented in Fig. 2(c). Finally, another PD produces a
new heterodyning process. In particular, the heterodyning pro-
cess of each replica with the closest optical carrier produces
the down-converted replica of the received RF signals in the
starting IF regions. At this point, the signals are filtered by a
low-pass filter (LPF), see Fig. 2(f). Then, they are sent to a 400
MSample/s analog-to-digital converter (ADC), which is finally
used to digitize both the received and the reference (i.e., the
transmitted RF signal) signals for the processing, by means of
a digital signal processing (DSP) unit.
III. D
IFFERENTIAL PHASE ESTIMATION
It is possible to precisely determine the target displacement
by exploiting a combination of SFCW signal modulation and
differential phase measurements [10]. As already mentioned, a
radar system which employs SFCW signal modulation trans-
mits a sequence of consecutive coherent sinusoidal signals. The
frequency of the mth generic sinusoid is given by:
f
m
= f
0
+(m − 1)Δf (1)
with m =1, ..., N
step
being the step index, where N
step
is the
number of sinusoids, Δf is the constant frequency separation,
namely the frequency step, and f
0
the carrier frequency, which
can assume both the values CF
S
and CF
X
. Finally, the total
signal bandwidth BW = N
step
Δf defines the achievable range
resolution ΔR = c/(2 · BW).
As known, a variation in the target distance will result in a
frequency dependent phase shift of the radar echo. By comparing
Fig. 2. Scheme of the photonics-based radar system (left), and spectrum in the highlighted point (right). (A) optical spectrum of the mode-locked (MLL)
laser used as optical local oscillator; (B) electrical spectrum of the applied radar signals; (C) optical spectrum of the optical modulated signal after the MZM;
(D) electrical spectrum at the photodiodes (PD) output; (E) electrical spectrum of the filtered RF radar signals before the wideband antenna; (F) electrical spectrum
of the received signals radar at intermediate frequency (IF), after the receiver PD.
the frequency phase of two consecutive echo acquisitions it is
possible to determine the target displacement. In fact, each mth
frequency component of the SFCW signal accumulates, during
the propagation, a phase shift Φ
m
proportional to the traveled
radial distance. Let us suppose that the radar system transmits at
time t
0
, and that the echo signal from a steady target at distance
d is collected by the receiver at time t
0
+ τ . Supposing that
the propagating medium is just air, the mth sinusoidal signal at
frequency f
m
will accumulate a relative phase contribution Φ
m
given by:
Φ
m
=2π · f
m
· τ =2π · f
m
·
2d
c
(2)
Thus, comparing two consecutive acquisitions of the same mth
harmonic, any change Δd
m
in the target distance, can be esti-
mated by the signal phase variation ΔΦ
m
:
Δd
m
=
c
4π · f
m
· ΔΦ
m
(3)
If the phase varies more than 2π, displacements ambiguities
come out, while the maximum unambiguous range R
ua
, can be
evaluated as:
R
ua
=
c
2 · Δf
(4)
By considering (1) and (3), we can observe that a given precision
in the phase estimation corresponds to a different accuracy in the
range determination, depending on Δf. The larger is Δf,the
more precise the displacement measure, at the cost of reducing
the unambiguous range [13].
If several targets are present in the scene under analysis,
and they belong to distinct range cells (i.e., their distance is
larger than ΔR), we can use standard radar algorithms to pro-
cess the data. Most implementations use the Inverse Discrete
Fourier Transform (IDFT) to transform the data into the spa-
tial domain [14]. By applying proper decomposition algorithms
to the received signal, it is possible to determine the separate
displacement of each target [15].
Furthermore, as pointed out in [13], the use of large Δf
allows to detect displacements much smaller than a fraction of
millimeter. In our case, higher displacement resolution can be
obtained by exploiting the dual-band radar operation. With our
setup, it is possible to obtain N
step
independent measures with
a maximum frequency difference Δf up to 7.4 GHz (i.e., the
difference between the two selected carriers). Moreover, since
the unambiguous displacement is limited to approximately 2 cm,
other couples of tones with different Δf can be eventually used
to resolve ambiguities and to provide a second stage of more
precise range evaluation. In this sense, further information can
be found in [10], [12].
IV. S
IMULATIONS
Numerical simulations were carried out in order to prelimi-
narily demonstrate the effectiveness of the proposed solution.
Setup parameters were chosen accordingly to our laboratory
equipment availability. In particular, the front-end filters band-
width is currently limited to 20 MHz, thus forcing us to limit
the SFCW signal bandwidth to BW ≤ 20 MHz.
By imposing the maximum unambiguous range R
ua
equal to
500 m, it is possible to retrieve from (4) the frequency step Δf ,
which is the analog of the pulse repetition frequency (PRF) in
pulse radar nomenclature [2]. In this way, given the total system
bandwidth limited to BW = 20 MHz, the number of frequency
steps can be retrieved as N
step
= BW/Δf = 66. Thus, the
maximum unambiguous range R
ua
is divided into N
step
range
cells with extension ΔR. In order to obtain a range resolution
ΔR = 7.5 m, the duration T
step
of each step must be equal to
50 ns, in other words T
step
=1/BW . These parameters were
set common to both the carrier frequencies, resulting in two
identical coherent sequences of SFCW signals being transmitted
in the S- and X-bands.
All the radar setup parameters are summarized in Table I,
while five targets were simulated simultaneously at 25, 145,
220, 300 and 400 m distance from the radar. As already said, in
our case for simplicity, but without loss of generality each target
is in a different range cell.
Each target is characterized by its own arbitrary displacement
pattern, as depicted in Fig. 3. However, in a real scenario the
displacement independence assumption among confining range
cells will be not necessarily true. Let us think, for instance, to the
TABLE I
S
ETUP PARAMETERS OF THE S- AND X-BAND SFCW SIGNALS TRANSMITTED
BY THE
PHOTONICS-BASED RADAR SYSTEM
Fig. 3. Simulated displacement curves versus the data record index. Five
targets are considered at 25 (red), 145 (yellow), 220 (purple), 300 (green) and
400 m (cyan) distance from the radar, respectively. Displacements are assumed
to be independent.
structural movements along the side of a mountain. Moreover,
different values of radar cross section (RCS) values were con-
sidered to simulate real and different targets (e.g., walls, slopes,
bridges, etc.).
The processing chain undergone by the received IF signals
s(t), respectively at IF
S
and IF
X
frequencies [see Fig. 2(f)],
where t denotes the time dependence, is illustrated Fig. 4. How-
ever, in the proposed scheme, the dependence on the carrier
frequency is omitted for brevity.
First, the cross-correlation is carried out between the received
signal s(t) and the reference signal r(t) to evaluate the presence
of possible targets and their distance. In the plot on the left
of Fig. 4 the obtained cross-correlations are shown for the five
simulated targets, exactly centered at their respective range dis-
tances. At the same time, the mixed signal is evaluated in the
time domain as m(t)=r(t) · s
∗
(t), where
∗
is the complex con-
jugate operator. From the analysis of the cross-correlation peaks,
once the presence of a target and its distance from the radar are
determined, the phase of M (f), i.e., the Discrete Fourier Trans-
form (DFT) of m(t), is evaluated for those spectral components
corresponding to the estimated ranges. M(f) is depicted in
the right plot of Fig. 4. By comparing two successive phase
measurements, the target displacement can be determined by
applying (3).
The simulated results are depicted in Fig. 5, which shows
the error curves obtained from the true and the estimated dis-
placements for the five simulated targets in the S-band. In the
simulations, fifteen sets of measurements are generated, each
consisting of 1000 pulse repetitions of the SFCW signal. The
first record is taken as a reference for the successive phase cal-
culations. The indexes of the data records are plotted along the
abscissas in Fig. 5. As we can observe, the estimation error is
usually smaller, in modulus, than 0.2 mm for all the five simu-
lated targets up to 400 m. However, it is worthy to observe that
the estimated displacements are also slightly influenced by the
relative positions of all the targets in the scene. This is clear for
targets 4 and 5, at 300 and 400 m and depicted by the green and
blue curves respectively, for which some degree of correlation
can be observed between the two.
A further analysis is conducted considering different values of
signal-to-noise ratio (SNR) affecting the received signal. With-
out loss of generality, with the term noise we can indicate both
the system noise and the clutter collected from the scene under
analysis. In this study, a complex white Gaussian noise has been
considered for simplicity.
The results of the numerical simulations are shown in Fig. 6.
The error curves between the true and the estimated displace-
ment are estimated only f or the second target (i.e., the one at
145 m). As for the previous analysis, fifteen sets of measure-
ments are generated, each composed of 1000 pulse repetitions
of the SFCW signal. The real target displacement behavior is
depicted by the ochre curve in Fig. 3. The final evaluated dis-
placement is estimated through differential phase measurements
between the S- and X-bands, therefore, considering a larger Δf
of 7.4 GHz between the two S- and X-band carriers [10]. It
can be observed that the absolute mean error is well below the
millimeter threshold, for positive SNR values larger than about
10 dB. Below this value, the error rapidly increases because
the useful signal becomes even more submerged in noise. This
event is comparable to the situation in which the target RCS is
smaller than the noise RCS. A possible solution to avoid this
problem could be to increase the radar range resolution capabil-
ity (i.e., decrease ΔR) and have a finer angular resolution (e.g.,
using a more directive antenna or synthesizing a more directive
beamwidth through SAR processing). At the current moment,
these possibilities are actually under further investigation.
V. E
XPERIMENTAL RESULTS
A preliminary experiment was conducted to prove the effec-
tiveness of the photonics-based transceiver for detecting and
measuring displacements when multiple targets are present. In
the experiments the same radar system parameters reported in
Table I were used. The only exception was the length of each
acquisition record, which was set equal to 100 repetition periods
of the SFCW waveform. Two metal surfaces of about 0.25 m
2
were considered as targets. The first target was fixed at 25 m
away from the radar and it was used as a phase reference. The
second target, placed at 145 m, was moved with 1 mm steps from
0 to 100 mm. For completeness, the spectrum of the received
SFCW signal at the output of the ADC, and integrated over 1000
