IEEE Geoscience and Remote Sensing Magazine - December 2021 - 176

performance by means of an FL-GPR under realistic conditions.
Test facilities at Ingegneria dei Sistemi S.p.A., headquartered
in Italy [see Figure 1(b)], are equipped with a
moving platform that can support two or more antennas,
pointing toward a test site that comprises several resolution
cells of the FL-GPR system. The test field allows the inclusion
of heterogeneous soils. Data were gathered in the frequency
range from 0.4 to 2 GHz using a transmit and a receive horn
antenna, both connected to a network analyzer.
The antennas, spaced 93.5 cm apart and tilted at a 45°
angle, were mounted on the moving platform at a distance
of 1.42 m from the air-soil interface. The platform was
moved along straight tracks with a constant spatial step of
size 0.02 m. Scanning lines of about 8 m were used to collect
data over a sandy portion of the test site. The experiments
were performed after intense rainfall, which reproduced
challenging operational conditions. This resulted in a nonhomogeneous
background medium that consisted of two
layers: the upper layer, with thickness of a few centimeters,
was dry sand while the deeper layer comprised wet sand.
This work also discussed the performance achievable with
a conventional microwave tomographic approach to focus
FL-GPR data.
NUMERICAL APPROACHES
Numerical data obtained by means of a finite- difference
time- domain (FDTD) method, modeling an FL-GPR system
Tx: (θinc, φ1)
Tx: (θinc, φ2)
Rx Array
4
3
2
1
-1
-2
-3
-4
1
7
4
-8 -6 -4 -2 02468
x (m)
(a)
4
3
2
1
-1
-2
-3
-4
-8 -6 -4 -2 02468
x (m)
(b)
FIGURE 2. The focused numerical data (in decibels) for a scene with size equal to 9 × 19 m:
ground with (a) a flat surface and (b) a randomly rough surface characterized by a root mean
square (rms) surface height equal to 0.8 cm and correlation length of 14.93 cm. Further details
can be found in [38]. Rx: receive; Tx: transmit. (Source: [38].)
176
-25
-30
-35
-40
-45
-50
-55
with two transmitters and 16 receivers, were presented in [18].
The modeled system has a frequency bandwidth from 0.3
to 1.5 GHz. In particular, a near-field army FDTD software
package was developed at the ARL for synthesizing FL-GPR
numerical data accounting for realistic sensing conditions.
More information on the modeling approach and targets
used for the analysis can be found in [37] and [38].
An example of FL-GPR-focused numerical data from [38]
is reproduced in Figure 2; the images are formed over a
horizontal plane in front of the transmitting and receiving
antennas considering both metallic and plastic targets
whose locations are specified in Figure 2(a). Both
the flat ground interface [Figure 2(a)] and the rough
surface [Figure 2(b)] are simulated. The latter was generated
by assuming a random process model described by
Gaussian statistics.
A 3D full-wave approach, based on a finite- difference
frequency- domain method and optimized to provide realtime
solutions, was proposed in [20] to model an FL-GPR
on a moving platform and calculate the scattering from
rough terrains located at large electrical distances from the
antennas. For a synthetic aperture, the computational domain
was reduced to a small subset of the observed region,
and the surface clutter was determined by performing the
simple multiplication of a precomputed impulse response
matrix of the rough profile with a matrix characterizing the
FL-GPR transmitted signal.
This approach significantly re5
2
11
3
6
9
10
8
-20
-25
-30
-35
-40
-45
-50
duced
the complexity through an efficient
use of computational resources,
thereby permitting the representation
of lossy/frequency-dispersive
soils and target-detection processing
in real time. This is especially useful
in scenarios where an experimental
performance validation may incur a
high cost and/or require significantly
more resources.
The authors in [39] extended the real-time
3D simulation to the multiview
case, considering a realistic velocity of
the moving platform. The matrix-multiplication-based
surface clutter computation
in this case required an additional
precomputed correction matrix
of the moving platform measurement
steps along the direction of motion.
The method was tested via Monte Carlo
simulations. In practice, the proposed
simulation-based approach can be
used to estimate the scattering from the
rough surface profile, which can then
be subtracted from the actual FL-GPR
measurements. The resulting difference
signals can then be processed for image
formation and target detection.
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE DECEMBER 2021
y (m)
y (m)

IEEE Geoscience and Remote Sensing Magazine - December 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - December 2021