IEEE Geoscience and Remote Sensing Magazine - December 2021 - 185

developed by PSI was proposed in [85]. The method relied
on the definition of a set of spatial lanes in the radar
image. The identification of potential targets was first independently
performed in each subregion, and they were
then tracked across the subregions. Weighted averages of
the corresponding geometrical features were evaluated,
and the target persistence across the spatial regions was
used to reduce the false-alarm rates. Targets appearing in
a limited number of lanes were removed as part of the detection
scheme.
An analysis of the spectral features extracted from the
scattered signal, with the goal to improve the performance
of buried explosive hazard detection, was provided in [86].
Natural resonant frequency and polarization features of
improvised explosive devices were examined in [87] for FLGPR.
In [76], a spectrum-based classifier was proposed that
rejected false alarms by classifying each potential target
based on its spatial frequency spectrum.
A method based on the use of narrow-band and fullband
radar processing, coupled with a classifier exploiting
complex-valued Gabor filter responses, was proposed in
[66]. Full-band radar images yielded high spatial resolution,
while narrow-band images provided the means to detect
targets with unique signatures. A composite confidence
map was implemented to detect local maxima and isolate
potential target pixels.
FUTURE TRENDS
A completely different radar-based approach from FL-GPR
to road mapping and clearing is to employ a traditional airborne,
side-looking SAR system flying on a track parallel to
the road and imaging the ground area of interest. This approach
has the advantages of a high coverage rate as well as
the fact that the platform does not come in contact with the
in-road hazard. Nevertheless, these radar systems operate
at relatively long ranges (at least 1 km) and, consequently,
require larger transmitted power and longer coherent integration
intervals to achieve high image resolution. Both
modeling and experimental studies have demonstrated the
difficulty of detecting weak buried targets (such as plastic
landmines) by side-looking GPR systems, even in mild clutter
conditions [21], [88].
Recent advances in radar sensors based on unmanned
aerial vehicle (UAV) platforms promise to bring together
the advantages of both types of aforementioned imaging
systems [89]-[96]. Thus, a UAV-based SAR system can
operate at small elevations and ranges, requiring a small
amount of power and a short synthetic aperture length.
At the same time, as the flying platform does not come in
contact with the ground, the standoff range requirement
relevant to ground-vehicle-borne systems does not apply
in this case.
A UAV-mounted radar system is likely to be significantly
less costly to build and operate than any of the current
airborne or FL-GPR systems, while the excellent control of
UAV flight trajectories makes motion compensation easier
DECEMBER 2021 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
to accomplish. The radar antenna can be readily configured
as down-looking, side-looking, or FL, depending on the imaging
application, while antenna arrays can be combined
with the synthetic aperture created by platform motion.
Another possible scenario is
using a network of distributed
UAV-based SAR imaging sensors
working cooperatively to
map a ground area. Vehicleborne
radar systems will still
have a role in explosive hazard
detection; however, we envision
that future trends will
favor employing unmanned
ground vehicle platforms for
this application.
One potential capability
ONE POTENTIAL
CAPABILITY OF FL-GPR
SYSTEMS THAT HAS RARELY
BEEN EXPLOITED IN
PRACTICE THUS FAR IS THE
CREATION OF 3D IMAGES OF
THE SCENE UNDER
INVESTIGATION.
of FL-GPR systems that has
rarely been exploited in practice thus far is the creation of
3D images of the scene under investigation. By adding the
height dimension to a radar map, one can infer the depth
of a buried target and partially mitigate the surface clutter,
which now affects only a limited part of the image volume.
Additional target features inferred from the 3D map can
also be useful in target-classification applications.
One example of a GPR system operating on the FL SAR
principle and capable of creating 3D images is the iRadar,
developed by the Lawrence Livermore National Laboratory
[97]. While the iRadar antenna array is mounted close
to the ground and provides only a modest standoff range,
one can envision a system equipped with a similar array
mounted on a UAV flying at a height of 1-2 m over the
road and mapping the area of interest, including the underground
volume [96]. Although the surface clutter becomes
less an issue in the detection of buried targets in
3D images, underground inhomogeneities created by different
soil layers, rocks, roots, and so on represent a new
source of clutter that may degrade the detection and classification
performance.
In addition to buried object detection, FL radar technology
is finding use in other emerging applications. One example
is attempting to exploit the 3D imaging capability of
an FL radar to assist helicopter landing in degraded visual
environments (DVEs), such as those created by brownout
conditions. A prototype radar system based on this principle
is currently being developed at ARL [98].
To achieve an angular resolution of 0.1-0.2°, comparable
to optical sensors such as lidar, this radar system must
operate in the millimeter-wave regime (Ka band). The wave
attenuation through dust, sand storms, or other DVE conditions
at these frequencies (less than 1 dB/km one way) is
still low enough to provide a see-through capability, which
is not available in infrared, optical, or lidar sensors. The 3D
map of the landing zone obtained by the FL SAR would be
interpreted in terms of natural or human-made terrain features,
and this information would be passed to the pilot via
185
IEEE Geoscience and Remote Sensing Magazine - December 2021

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