Professional Documents
Culture Documents
SPIEDigitalLibrary.org/conference-proceedings-of-spie
ABSTRACT
Through wall indoor surveillance and monitoring is an emerging area of research and developments in many real-life
applications, including security and search and rescue. In this context, it is important to deploy remote sensors able to
glean information about the presence of people in indoor areas. Radar devices based on the exploitation of
electromagnetic signals represent an attractive sensing modality. When operating at frequencies from hundreds of MHz
to few GHz, they are capable to penetrate common construction materials enabling the detection of people in
inaccessible environments. Two different types of surveillance radars are considered in this paper. The first one exploits
the synthetic aperture concept with the aim to detect, locate, and track multiple subjects in the scene. The second one
illuminates the scene from a fixed position and exploits the Doppler effect. Despite the limited information available
with respect to aperture radars, Doppler radars allow the detection of moving targets behind walls and inside enclosed
structures. This work presents a brief overview of recent developments in the field of radar data processing for indoor
monitoring together with their assessment by means of numerical and experimental tests.
1. INTRODUCTION
Radar imaging is a broad research topic underpinning any sensing operation seeking detection, localization and tracking
of targets, including gleaning knowledge of target response to an incident electromagnetic signal [1]. Therefore, radar
imaging using remote sensing is key to indoor surveillance where humans represent the target of interest.
The primary goal of radar imaging is to obtain information about the target location and geometry by means of an
aperture radar (either synthetic or physical). The radar illuminates the scene and collects the signals scattered by the
target. This goal involves applying specific data processing algorithms that process the input raw radar backscattering
signals and produce an output image of the scene. This image should be easily interpretable, permitting target
recognition and characterization.
The radar imaging literature is extensive and provides numerous signal processing algorithms, which have appeared in
different applicative contexts such as Synthetic Aperture Radar (SAR) [2], Through-Wall Radar (TWR) [3], Ground
Penetrating Radar (GPR) [4], biomedical imaging [5], etc.
The imaging or focusing algorithms can be broadly classified into beamforming, or matched filtering, and inverse
filtering methods [6]. Beamforming schemes [7]-[10] perform imaging by compensating the phase or the delay of the
scattered field with a proper spatially varying filter, similar to the approaches applied in seismic exploration. On the
other hand, inverse filtering methods are more rigorous in that they are based on the electromagnetic (EM) modeling of
the wave scattering phenomenon, and invert the mathematical relationship between the scattered field data and the
unknown EM properties of the target [11]-[12]. This relationship depends on the target nature (dielectric, metallic), the
radar imaging configuration, and the operating scenario. Additionally, the inverse filtering methods are characterized by
a large flexibility in dealing with unconventional measurement configurations, such as multi-bistatic and multi-
view/multi-static sensor deployment, as well as scenarios different from the free-space assumption. Furthermore, sparse
reconstruction approaches based on the compressive sensing paradigm have recently attracted considerable attention due
to their superior super resolution performance and ability to deal with incomplete data measurements [13].
2. RADAR IMAGING
This Section presents recent advances in the indoor surveillance in relation to state of art radar imaging approaches. It is
divided into subsections relative to different scenarios arising in the context of indoor surveillance.
The 2D geometry depicted in Fig. 1 is important to present the radar imaging problem from a general perspective. As can
be seen, an indoor scene comprising stationary and moving targets is probed by an aperture array (synthetic or physical)
of transmitting (Tx) and receiving (Rx) antennas. At any measurement point, the Tx antenna emits a wideband pulse in
the radar frequency band and the Rx antenna records the signal scattered by the targets. When the Tx and Rx antennas
are located at the same point or have a negligible spatial offset in terms of wavelength, the measurement configuration is
denoted as multimonostatic/multifrequency. A more sophisticated configuration is the multiview/
multistatic/multifrequency or Multiple Input Multiple Output (MIMO), where Tx and Rx are located at different
positions and are moved independently of each other.
Once the radar signals have been collected, a suitable data processing is needed to extract pertinent information (target
detection, position, size, etc.). It is necessary to process the data so as to filter unwanted clutter signals that are not
produced by the targets of interest (e.g. direct coupling from Tx to Rx, wall reflections, multipath, etc.). It is also
necessary to process the data in such way to obtain well-focused images that reproduce the scene accurately and reliably.
Regardless of the mathematical formulation of the sensing phenomenon, the imaging task amounts to determine an
unknown function accounting for target reflectivity from scattered field data. These aspects will be discussed in the
following subsections in relation to the radar imaging configurations that are relevant to indoor surveillance.
Common approaches to filter front-wall reflections are the background removal [3], the spatial filtering [22], and the
subspace wall projections [23]. A comparison between different front-wall clutter rejection methods has been reported in
[24].
A large amount of research work has been done on focusing algorithms. The most popular class of imaging approaches
exploits the beamforming strategy [7]-[10], which represents the first attempt towards TWRI. Basically, the
beamforming determines the complex image value at any point in the scene by passing the received signal through a
filter matched to the transmitted pulse.
Another important class of TWRI approaches is based on the EM inverse scattering methods such as the non-linear
tomographic approach [25] and the linear inverse scattering techniques based on the diffraction tomography [26] and
Microwave Tomography (MT) [27], [28]. The aforementioned imaging methods hold for homogenous slab walls, and
their effectiveness has been experimentally assessed. The extension of linear inverse scattering approaches to more
complicated scenarios involving inhomogeneous cinderblock walls [29], [30] and building corners [31] has been
reported as well.
Former studies highlight that an accurate modeling of the EM scattering phenomenon is necessary to obtain images free
from artifacts and distortions. One issue affecting the reliability of TWRI is the inaccurate information of the
electromagnetic properties of the wall. Therefore, signal-processing strategies addressing this issue have been developed.
An autofocusing technique, based on higher order statistics, was presented to correct for errors under unknown walls
[32]. A simple procedure to estimate the wall transmission coefficient for symmetric lossless walls was introduced and
validated by synthetic data in [33]. A two-step imaging procedure was presented in [28], where the thickness and the
dielectric permittivity of the wall were first estimated taking into account that actual measurements (i.e., the field
reflected by the wall plus the one scattered by the obscured scatterers). The problem was then cast as a linear inverse
scattering one and solved by means of the Truncated Singular Value Decomposition algorithm. The high-frequency EM
characterization of through-wall building imaging was presented in [34].
The imaging process is followed by higher-level signal processing operations. These include automatic target detection
[41], classification [42] and tracking [43]. The tracking of moving targets in a TW scenario has received little attention.
Indeed, the identification of moving targets has been mostly performed by change detection schemes based on the
coherent or incoherent subtraction of radar data/images [44]-[46]. However, target tracking is a complex and more
informative process that involves the automatic estimation of the dynamics (position and velocity) of one or more
targets. In this respect, a multiple extended target tracking procedure for real-time operation was reported in [43]. The
first stage of the technique regards a preprocessing of raw data and the image formation process, which is performed
according to a linear MT approach. Then, each tomographic image is given as input to an object detector, constituted by
a Constant False Alarm Rate (CFAR) detector followed by a segmentation procedure. The detected objects are processed
according to the Joint Probabilistic Data Association (JPDA) rule. The proposed tracking approach has been assessed in
single-target and multi-target TW scenarios pointing out how shadowing and mutual interactions effects can unavoidably
degrade the reliability of tracking results.
Fig. 4 Tomographic images obtained with the passive multi-array change detection procedure.
(a) (b)
Fig. 5 Architecture of a CW Doppler radar (a). Radar prototype developed at IREA-CNR (b).
A data processing strategy for occupancy sensing has been implemented based on the time varying analysis of the
standard deviation of the signal. Specifically, the standard deviation of the signal over a sliding window is compared to a
threshold value. If the standard deviation is higher than the threshold, the system generates a binary output equal to 1 to
indicate that someone is present in the scene. Conversely, if the standard deviation is lower than the threshold the system
produces an output equal to zero to inform that nobody is present. The threshold value is estimated from calibration
measurements and in particular from the standard deviation of the radar signal in the case of empty scene, since this
parameter provides a reasonable estimate of the system noise level.
An experiment was carried out in the indoor scenario shown in Fig. 6a featuring a corridor 20m long, and the radar was
installed at the beginning of it to detect the presence of human target. The duration of the radar acquisition was 180s. Fig.
6b shows that the raw radar signal exhibits strong fluctuations that are produced by the walking subject. The entity of
these fluctuations depends on the distance between the target and the radar. The numerical data plotted in Fig. 6c display
Figure 6. Radar installation in the corridor (a). Raw signal (b) and detection output (c).
4. CONCLUSIONS
This paper provided a short overview of theoretical and practical aspects of radar imaging and in particular of data
processing algorithms. These algorithms provide an image of the scene under investigation, which is informative and
much more easily interpretable than the raw data. Radar imaging systems specific to security applications such as
through-wall radar and passive multistatic radars were presented. The paper reported numerical/experimental tests to
demonstrate radar operations and highlight their imaging potentials. Moreover, radar devices based on the exploitation of
Doppler information were presented as a low-cost effective tool for indoor surveillance. These sensors provide a simple
information for situation awareness of the scene, yet they are very suited for integration and fusion into multimodal
surveillance systems comprising different sensing technologies (RGB cameras, thermal cameras, etc.) to enable
enhanced surveillance.
REFERENCES
[1] Cheney, M. and Borden, B., [Fundamentals of Radar Imaging], vol. 79, Siam, (2009).
[2] Cumming, I. G., and Wong, F. H., [Digital Signal Processing of Synthetic Aperture Radar Data: Algorithms and
Implementation], Artech House, (2004).
[3] Amin, M. G. [Through-the-Wall Radar Imaging], CRC Press, Boca Raton, FL, USA, (2011).
[4] Persico, R., [Introduction to Ground Penetrating Radar: Inverse Scattering and Data Processing], John Wiley &
Sons, (2014).
[5] Nikolova, N. K., “Microwave imaging for breast cancer,” IEEE Microwave Magazine 12 (7), 78–94, (2011).