You are on page 1of 5

2016 IEEE Radar Conference (RadarConf)

Radar Processing Architecture for Simultaneous


SAR, GMTI, ATR, and Tracking
Ryan K. Hersey

Edwin Culpepper

Sensors and Electromagnetic Applications Laboratory


Georgia Tech Research Institute
Atlanta, GA

Sensors Directorate
US Air Force Research Laboratory
Wright-Patterson AFB, OH

AbstractCombined synthetic aperture radar (SAR) and


ground moving target indication (GMTI) radar modes
simultaneously generate SAR and GMTI products from the
same radar data. Furthermore, the SAR and GMTI data
products can be further exploited for target signature
extractions, automatic target recognition (ATR), and featureaided tracking. This hybrid mode provides the benefit of fused
imaging and moving target displays along with enhanced target
recognition and ground target tracking.
The Air Force
Research Laboratory (AFRL) Gotcha radar has collected widebandwidth, multi-channel data that can be utilized for these
hybrid mode applications. This paper presents a processing
architecture for simultaneous SAR, GMTI, ATR, and tracking,
and includes the results of applying this processing to the AFRL
Gotcha data.
Keywordsradar; signal processing; SAR; GMTI; ATR;
tracking

I.

INTRODUCTION

Combined synthetic aperture radar (SAR) and ground


moving target indication (GMTI) radar modes simultaneously
generate SAR and GMTI products from the same radar data.
This combined capability provides many benefit including the
ability to combine radar imagery and moving target displays,
allowing for improved target detection, tracking, and
recognition. However, the differing system, antenna, and
waveform requirements between SAR and GMTI modes make
implementing the hybrid mode challenging. SAR systems
typically use long dwells and high bandwidths to generate a
fine-resolution image of stationary ground terrain. GMTI
systems use short dwells and low bandwidths to find moving
targets over a wide area. The combined implementation of
SAR and GMTI often requires comprises in the individual
mode performance.
The AFRL Gotcha radar has acquired persistent
surveillance, wideband radar data over a suburban
environment. This data has been used to demonstrate SAR,
change detection, and GMTI radar modes [1-7]. This paper
expands many of these processing concepts into an integrated
architecture incorporating layered operator displays and
interfaces. The integrated architecture combines the principles
of SAR, GMTI, target recognition and tracking into a common
framework with shared data products. The integrated displays

978-1-5090-0863-6/16/$31.00 2016 IEEE

present this information to an operator or analyst using layered


displays of data products with both wide-area and small-area
views.
This paper is organized as follows. Section II presents the
radar processing architecture including descriptions and
integrations of the processing stages. Section III presents
results of the radar processing architecture with Gotcha data.
Section IV summarizes the paper.
II.

RADAR PROCESSING ARCHITECTURE

Figure 1 shows the combined radar processing architecture


incorporating essential components of SAR, GMTI, ATR, and
tracking. This architecture integrates the following six radar
signal processing stages: SAR polar formatting, SAR
backprojection, sub-banded GMTI, target signature extraction,
ATR, and feature-aided tracking. The output SAR imagery,
GMTI detections, target classifications, target tracks, and
target signatures are displayed using an integrated layered
display that fuses the multiple data products.
A.

SAR Polar Formatting


The SAR polar formatting stage applies the polar format
algorithm to data from all spatial channels. This processing
step serves three primary functions in the architecture. First,
SAR polar formatting generates wide-area SAR imagery for
operator displays. Second, SAR polar formatting serves as a
focusing stage for GMTI processing, correcting for target
range migration. Finally, SAR polar formatting provides an
initial focusing stage for signature extraction from both
stationary and moving targets. Other SAR imaging algorithms
such as the keystone algorithm or the range-migration
algorithm may be applied instead of the polar format
algorithm.
B. SAR Backprojection
SAR backprojection provides fine-resolution imagery for
operator displays and signature extraction from stationary
targets. Backprojection is nominally applied to either a single
channel or a combination of channels through digital
beamforming thereby generating a single backprojection
image.
Multi-channel backprojection images could be
generated if desired.

Figure 1 Radar processing architecture for simultaneous SAR, GMTI, ATR, and tracking

The persistent nature of the AFRL Gotcha collection


allows for the generation of video SAR data, where the
backprojection SAR image is updated at a specified intervals
of time. The backprojection video SAR processing forms subaperture images at the desired update rate. The sub-aperture
images are then coherently combined to form the full
resolution image.
C. Sub-Banded GMTI
The architecture utilizes a sub-banded, space-time adaptive
processing (STAP) approach to detect and geolocate moving
targets in the wideband, multi-channel radar data. Details of
this approach are described in further detail in [1].
Sub-banding provides two primary benefits to GMTI
processing. First GMTI often performs better on narrow-band
data as targets are less likely to be sub-resolved or move
through range resolution cells. Second, since each sub-band
has a different center frequency, the grating lobes for each
sub-band appear at different angles. The method uses this
variation to disambiguate target returns and places them at the
correct AOA, which is a critical and challenging problem to
overcome for the Gotcha collection.

D. Signature Extraction
The radar processing architecture can extract radar
signatures for both stationary and moving targets of interest.
The targets of interest will nominally be selected by a user, but
this process could be automated with additional logic and
sensor management.
Signatures from stationary targets are extracted primarily
from the single-channel backprojection imagery.
The
persistent surveillance of the Gotcha collection allows for
signatures to be collected from a variety of aspect angles.
Signatures from moving targets are extracted primarily
from the multi-channel polar format imagery.
GMTI
detection information (range and Doppler) is needed to
properly locate the moving-target signatures in the data. The
GMTI detection information can be augmented with GMTI
track information though an optional feedback loop. After the
multi-channel signature data has been extracted, STAP clutter
mitigation and inverse SAR (ISAR) focusing may be applied
to improve the target signature quality.
E. ATR
The radar processing architecture can apply ATR to the
extract target signatures. ATR could be applied to a variety of
target signature data including SAR imagery of stationary

targets, high range resolution (HRR) profiles of moving


targets, and ISAR imagery of moving targets. A target library
would be established via offline training.
F. Feature-Aided Tracking
Tracking of vehicle and dismount targets in complex
environments can be problematic due to the challenges of
associating detections with the correct tracks. Common
GMTI tracking problems include tracking move-stop-move
targets, tracking targets through turns, and tracking in dense
target environments.
Target features derived from the target classification
and/or the target signatures can help associate detections with
tracks. The radar processing architecture supplies target
signatures and target classifications along with GMTI
detections to a feature aided tracker. The feature-aided tracker
will generate target tracks that can be both displayed to the
user and applied in a feedback manner to enhance target
signature extraction.
III.

RESULLTS WITH GOTCHA DATA

The AFRL Gotcha data presented here has 3 spatial


channels of covers a 1.25 km ground spot. The Gotcha 1
system operates at X-band with a 640 MHz bandwidth. The
Gotcha 1 collection was designed primarily as a SAR /
coherent change detection (CCD) system; however, the three
spatial channels allow some GMTI capability. The large

clutter Doppler spread in the Gotcha data makes detecting


slow-moving targets difficult. Additionally, grating lobes
created by the sparse receive channel separation make target
geolocation and false-alarm mitigation challenging.
Moving targets displace in cross range from their true
location in the SAR image. An exo-clutter target is a target
that moves fast enough to compete with only sidelobe clutter.
An endo-clutter target moves slowly such that it compete
with mainlobe clutter. Exo-clutter targets can usually be
detected using simple Doppler processing, whereas endoclutter targets require more advanced clutter mitigation such
as STAP [8-10]. Figure 2 shows a Gotcha SAR image with
indicated target detections. The figure shows both the true
and displaced detection locations. The figure also shows both
vehicle and dismount (human) targets, which are
discriminated using radial velocity and signal-to-noise ratio
(SNR) thresholds. The detection indicated with the yellow
ellipses is an exo-clutter target. The displaced detection
signature is clearly visible in the SAR image and can be
detected using a simple constant false alarm rate (CFAR)
detector. The detections indicated with the green and
magenta ellipses are endo-clutter targets. For these detection,
the displaced signature is masked by strong mainlobe ground
clutter. STAP processing is able to detect this slow-moving
target as shown in the clutter-mitigated image. After clutter
mitigation, several additional targets can be detected despite
competing with mainlobe clutter.

Figure 2 SAR and STAP processing of Gotcha Data

Figure 3 shows an integrated SAR-GMTI display of Gotcha


data for both wide-area and small-area views. This display
shows a background optical image with an overlay of the
SAR backprojection image and GMTI detections. The
backprojection SAR image aligns very well with the
background image. The GMTI vehicle detections (magenta
and yellow dots) generally appear on roads as expected.
Additionally the vehicle detections align very well with the
vehicle truth data (blue dots) for the two moving vehicles.
Note that there are several parked vehicles in the scene that
are not detected. The majority of the dismount detections
(red and green dots) appear in the field near the dismount
truth (cyan dots).

Figure 4 shows an enhanced version of Figure 3 that


includes the following displays:
Wide-area, SAR-GMTI view
Small-area, SAR-GMTI view
Target SAR signature
Target STAP signature
The wide-area, SAR-GMTI view provides a layered
display of the full 1.25 km Gotcha spot. This layers in this
view include the following:
Background optical image
Wide-area backprojection SAR image
Wide-area GMTI vehicle and dismount detections
and/or tracks
Coverage area (solid white lines)
Small area selection (solid white rectangle)
Antenna beam location (dashed white line)
This wide-area view gives the operator general situational
awareness as well as an interface to select targets of interest.
The small-area, SAR-GMTI view provides a layered
display of a user-selected target of interest. The user-selected
target could be a moving target that is cued by GMTI
detections or tracks, or the user-selected target could be a
fixed target in the SAR image. The size of the small-area
display in Figure 4 is 100 m. The layered display options in
the small-area, SAR-GMTI view are the same as the widearea, SAR-GMTI view.
The target SAR signature view shows the raw signature of
the user-selected target of interest. For moving targets, the
signature will be displaced in cross range. Furthermore, the
motion of the target causes the target signature to defocus in
cross range as is the case in the example in Figure 4. ISARbased focusing algorithms could be applied to improve the
quality of the signature of this moving target.

Figure 3 Integrated optical, SAR, and GMTI display of Gotcha data with a
low-resolution, wide-area view (left) and high-resolution, small-area view
(right)

The target STAP signature view shows the target


signature after STAP-based clutter mitigation. The target
SAR signature of slower-moving, endo-clutter targets may be
masked by ground clutter. The application of STAP whitens
the ground clutter returns thereby improving that target
signal-to-interference-plus-noise ratio.

Figure 4 Integrated Gotcha wide-area and small-area displays including moving-target signatures with and without STAP clutter mitigation

IV.

SUMMARY

This paper presents an architecture for simultaneous SAR,


GMTI, ATR, and tracking. This hybrid mode provides the
benefit of fused imaging and moving target displays along
with enhanced target recognition and ground target tracking.
The integrated architecture combines multiple processing
stages including SAR imaging, STAP-based sub-banded
GMTI, target signature extraction, ATR, and feature-aided
tracking. The integrated displays provide both wide-area and
small-area views of layered SAR imagery, GMTI detection,
target signature, target classification, and tracking data
products. This paper presents results using AFRLs Gotcha
data, including SAR, GMTI, and target signature data
products. Future efforts will further develop and test the
feature-aided tracking, target signature focusing and ATR
processing stages of the architecture.

REFERENCES
[1]

Hersey, R.K. and Culpepper, E., Sub-Band Processing for Grating


Lobe Disambiguation in Sparse Arrays, SPIE DSS 2014, May 2014.
[2] Page, D., Owirka, G., Nichols, H., Scarborough, S., Minardi, M., and
Gorham, L., "Detection and tracking of moving vehicles with Gotcha
radar systems," IEEE Aerospace and Electronic Systems Magazine, v
29, n 1, 50-60 (2014).
[3] Guo, B., Vu, D., Xu, L., Xue, M., and Li, J., Ground moving target
indication via multichannel airborne SAR, IEEE Transactions on
Geoscience and Remote Sensing, v 40, n 10, p 3753-64 (2011).
[4] Demming, R., Ilin, R., and Best, M., "Phase-wrapping ambiguity in
along-track interferometry," Proc. SPIE 8746 (2013).
[5] Stojanovic, I., and Novak, L., "Change detection experiments using
Gotcha public release SAR data," Proc. SPIE 8746 (2013).
[6] Key Yong Li, Uysal, F., Pillai,S. U., and Moore, L. J., "Modified spacetime adaptive processing for dismount detection using synthetic
aperture radar," Proc. IEEE Radar Conference (2012).
[7] Demming, R., MacIntosh, S., and Best, M., "Three-channel processing
for improved geo-location performance in SAR-based GMTI
interferometry," Proc. SPIE 8394 (2012).
[8] Melvin, W. L., A STAP overview, IEEE AES Magazine, v 19, n 1, p
19-35 (2004)
[9] Guerci, J. R., Space-time adaptive processing for radar, Artech
House, Norwood, MA (2003).
[10] Klemm,
R., Space-time adaptive processing principles and
applications, IEE Radar, Sonar, Navigation and Avionics 9, IEE Press
(1998).

You might also like