Académique Documents
Professionnel Documents
Culture Documents
Abstract—A near-field radar imaging technique is presented transmit elements and nr receive elements located at rt and rr ,
based on single-frequency measurements from a multiple-input respectively. Here, the boldface symbols represent vector quan-
multiple-output (MIMO) array. Such arrays are able to fully sam- tities. Assuming the first Born approximation and isotropic
ple the k-space of a region of interest. Advantageously, both range
and cross-range information can be reconstructed, in contrast to antennas, the signal measured by each pair of transmit/receive
conventional single-input multiple-output systems which require elements can be expressed as:
wideband frequency illumination. A theory of single frequency Z
MIMO imaging is developed and studied through numerical s0 (rt , rr ) = g0 (rt , r) σ(r) g0 (r, rr ) d3 r, (1)
modeling and experimental validation. r
Index Terms—MIMO, Microwave imaging
where σ(r) is the target reflectivity and g0 (ra , rb ) is the
I. I NTRODUCTION free space Green’s function accounting for the propagation
of waves from ra to rb , with wavenumber k0 = 2πf0 /c.
Near-field radar imaging has found application in many By exploiting the spatial diversity that exists between the
fields, including medical diagnosis [1], through-wall imag- transmit and receive arrays and a frequency bandwidth, an
ing [2] and concealed weapon detection [3]. The resolution estimation of the target reflectivity can be computed on a
of these systems is usually facilitated by large and densely discrete spatial grid r by compensating for the impact of the
sampled antenna arrays that radiate ultra-wideband signals— Green functions. However, in the context of single-frequency
ensuring good coverage of the target space’s spatial Fourier processing, a comparison between a SIMO and a MIMO
domain (k-space). More recently, in an effort to develop more system demonstrates the advantage of exploiting both the
efficient imaging systems, the radar community has deployed spatial diversity in transmission and reception. To illustrate,
antenna arrays consisting of multiple independent transmission two identical one-dimensional arrays are studied (Fig. 1).
and reception elements to improve resolution, mitigate clutter,
and improve classification performance [4]. These multiple-
input multiple-output (MIMO) systems are built on exist- • •
• •
ing single-input multiple-output (SIMO) and multiple-input
• g0 (r •
single-output (MISO) solutions [4], [5], [6]. A MIMO system , rr g0 (r
) , rr
• • )
acquires measurements from independent transmission and •
g0 (rt , r )
y • y
reception antenna arrays. The principle presented in this paper • θt •
, r)
exploits the enhanced coverage of the k-space permitted by the • σ(r) • g 0 (r t σ(r)
use of these independent transmission and reception arrays. • •
of plane waves transmitted and received by the elements of (a) SIMO array. (b) Monostatic MIMO array.
each array, allowing for direct identification of the Fourier
Fig. 1: Radar imaging using a SIMO and a MIMO array. The
components measured by the imaging system [7]. We show
red triangles represent the transmit elements and the green
that if the target is in the near field of the array and a sufficient
circles, the receive elements.
number of transmitter and receiver positions are probed, these
measurement points can fully survey a target’s k-space—even In the depicted setup, the antenna locations are defined by
at a single frequency—obtaining both cross-range as well as xt and xr on the x-axis, allowing for an indexing of the
range information. The recovery of range information using measured signal as s0 (xt , xr ). A representation of the data
a MIMO array is in contrast with conventional radar systems in the Fourier space is given by computing the following two-
which must exploit a frequency bandwidth to fully sample the dimensional discrete Fourier transform:
k-space for depth resolution. XX
s0 (kxt , kxr ) = s0 (xt , xr ) ej(kxt xt +kxr xr ) (2)
II. S INGLE - FREQUENCY MIMO RADAR PROCESSING xt xr
To illustrate the technique of single-frequency MIMO imag- Considering uniformly sampled transmit and receive arrays
ing, we consider a generic antenna array consisting of nt of overall lengths ∆xt and ∆xr , of sampling δxt and δxr ,
ky
and centered on xt0 and xr0 , the antenna array locations are
defined as follows on the plane y = 0:
2k0
xt = {−∆xt /2 : δxt : ∆xt /2} + xt0 (3) ....................
...... ...
xr = {−∆xr /2 : δxr : ∆xr /2} + xr0 (4) .. ..
.. ..
. .
kr
allowing for the determination of the equivalent plane wave
components kxt and kxr : . .
kxt = {−π/δxt : 2π/∆xt : π/δxt } (5)
kxr = {−π/δxr : 2π/∆xr : π/δxr } (6) kt
kr
.. . ..
........... . ....... .. . . .. ................. . . . .
.... ........
++++++++++
... . . . ... ++
... .
. .....
++
++
... +
+
.. . ..
+
+
...
wave vectors. In a SIMO configuration, the use of a unique . +
+
kt . .
+
+
.
. .. .
+ +
kt
transmitter prevents the spatial control of the radiated field.
+ +
. . . kx
Following [9], the field radiated by a single transmit antenna −2k0
+ +
2k0
can thus be approximated around the target center location by
a plane wave kt defined by: Fig. 3: k-space coverage of a single-frequency MIMO imaging
system and two examples of transmit and receive plane wave
interaction.
kxt = k0 cos(θt ), kyt = k0 sin(θt ) (8)
with θt the angle of incidence defined relative to the x-axis
by the location of the transmit element and the center of the III. E FFICIENT NUMERICAL RECONSTRUCTION
target space (Fig. 1a). Thus, for single-frequency illumination,
A Stolt f - k migration is implemented to efficiently esti-
the dispersion relation is approximated as:
mate a target reflectivity from the measurement of a single-
frequency MIMO matrix, adapting the reconstruction tech-
q
ky (kxr ) = k0 sin(θt ) + k02 − kx2 r (9)
nique introduced in [5] to a single-frequency measurement.
The spatial frequency components in a single-frequency According to Eq. (2), the target reflectivity can then be
SIMO configuration thus lie on a curve, since the system has estimated by compensating for the phase of the plane waves:
no control over the direction of the transmitted plane waves
and the norm of the wave vectors is a constant (Fig. 2). XX
This limitation can be circumvented by implementing a σ̂(x, y) = s0 (kxt , kxr )
kxt kxr
MIMO setup. Given a particular size and sampling of the ar-
(10)
rays, defined by xt and xr as in Eqs. 3 and 4, the independent ejkxt (x−xt0 ) ejkxr (x−xr0 )
control over the transmit and receive plane waves allows for a √ 2 2 √ 2 2
ej ( k0 −kxt + k0 −kxr )y
wide-coverage of the k-space, defined by Eq. 7 and depicted
in Fig. (3). The spatial diversity is significantly enhanced This operation can be performed efficiently by interpolating
compared with the SIMO and MISO cases because of this the phased matched signal sM 0 (kxt , kxr ) on the mean target
supplementary degree of freedom. In the SIMO approach, location (xC , yC ):
the inter-element spacing, δxt and δxr , is c/(2f0 ) to avoid
jkxt (xC −xt0 ) jkxr (xC −xr0 )
aliasing effects and ensure maximum coverage of the k-space sM
0 (kxt , kxr ) =s0 (kxt kxr )e e
√ 2 2 √ 2 2
sampling, according to the dimensions of the radiating arrays. ej ( k0 −kxt + k0 −kxr )yC
Having identified the elements relative to the spatial diver- (11)
sity measured by a single-frequency MIMO radar system, we
next suggest an efficient reconstruction method and validate onto a uniform grid sI (kx , ky ) and finally estimating the
its performance with numerical simulations. target reflectivity with a two-dimensional inverse fast Fourier
1425
2017 11th European Conference on Antennas and Propagation (EUCAP)
-20
-30
-40
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
y (m)
Fig. 5: Signals extracted from the first line (top) and central
column (bottom) of the target array. The dashed black lines
correspond to the actual locations of the targets.
1426
2017 11th European Conference on Antennas and Propagation (EUCAP)
V. C ONCLUSION
A single-frequency MIMO imaging technique has been
presented that can resolve near-field objects in range and
cross-range. It exploits the independent control of the trans-
mitted and received plane waves available in MIMO antenna
array systems. The technique implemented in this paper is
compatible with real-time applications and might be applied
to antenna arrays connected to fast transmission and recep-
tion systems. By operating at a single frequency, hardware
and signal processing requirements are substantially reduced
Fig. 6: Experimental setup made of two translation stages as compared to conventional systems. The extrapolation of
allowing for the sequential acquisition of a single-frequency such a technique to three-dimensional imaging is possible
MIMO matrix to image an array of 5× 5 metal marbles. by using two-dimensional radiating apertures and modifying
the dispersion relation accordingly. Such an approach might
be of interest when applied to hyperspectral imaging, where
applied to the measurements, allowing for the reconstruction frequency selective and/or dispersive targets prevent the use of
of the scene’s reflectivity (Fig. 7). wideband coherent illumination. The idea of single frequency
near-field imaging can also be implemented using dynamic
0
metasurface antennas [11], [12], significantly simplifying the
Tx Antennas hardware used in this work. Such a device that is simple and
0.6 Rx Antennas fast in both hardware and software is the subject of our current
-5
0.4 work.
y (m)
1427