-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 341
L ’ in thearena of tactical battlefield surveillance beganin
1967 with a program to develop a radar sys-tem that would penetrate
jungle foliage and detectmoving hostile intruders. This effort
arose during thewar in Vietnam, when major national
laboratorieswere called upon to contribute solutions to
tacticalbattlefield surveillance involving ground-based
andairborne-based sensors.
Ground-based sensors can be loosely grouped intotwo categories.
Special ground-penetrating radar sen-sors are used to detect mines
and other explosives aswell as hidden tunnels and buried stores.
Otherground-based radar systems are used to survey largeregions of
terrain within the sensors’ fields of view inorder to detect and
identify fixed ground targets andto detect, identify, and track
moving ground targets.Airborne sensors designed for tactical
battlefield sur-veillance require the ability to survey large areas
onthe ground in a timely manner in order to detect andidentify both
fixed and moving surface targets thatmay be hidden in ground
clutter or protected bycountermeasures. Lincoln Laboratory has
developed a
Tactical Radars forGround SurveillanceThomas G. Bryant, Gerald
B. Morse, Leslie M. Novak, and John C. Henry
■ Battlefield awareness is the key to battlefield dominance. The
fieldcommander who knows the enemy’s location and the types of
forces beingdeployed enjoys a great tactical advantage. The problem
of detecting andclassifying ground targets presents substantial
technical challenges, whichLincoln Laboratory has addressed for
nearly three decades in its TacticalTechnology program. Substantial
progress has been made in many aspects ofground surveillance since
the mid-1960s, but many challenges remain. Thesechallenges include
sensor development, signal processing, and
target-recognitiontechnology. Among its successes, the Laboratory
has provided the foundationfor operational national assets such as
the Joint Surveillance Target Attack RadarSystem (Joint STARS)
airborne surveillance system. This article describes
inchronological order several important Laboratory tactical-radar
programs andthe technologies that were developed for both airborne
and ground-basedsurface surveillance.
broad understanding of the technology and the phe-nomenology of
target detection; the Laboratory hasalso developed a variety of
remote sensors, communi-cation strategies, digital processors,
signal-processingalgorithms, and data-processing techniques to
addressthe concerns of different surface-surveillance systems.This
article describes some of the significant Labora-tory radar
programs and the technologies that weredeveloped for these
ground-based and airborne sys-tems. For more information on these
and other pro-grams at the Laboratory, see the articles
entitled“Displaced-Phase-Center Antenna Technique,” byCharles
Edward Muehe and Melvin Labitt, and “De-velopment of Coherent Laser
Radar at Lincoln Labo-ratory,” by Alfred B. Gschwendtner and
William E.Keicher, both in this issue.
Foliage-Penetrating Radar (1967–1972)
Field reports from American troops in Vietnam re-vealed that
foliage played a major role in concealingthe enemy in most tactical
engagements. As a result,Lincoln Laboratory began an investigation
of a sur-veillance system that offered a foliage-penetration
ca-
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
342 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
pability. A preliminary Laboratory study in 1966concluded that
radar would be able to detect peopleand vehicles moving through
dense foliage.
The Camp Sentinel Radar
The foliage-penetration radar program, supportedoriginally by
the Defense Advanced Research ProjectsAgency (DARPA) and
subsequently by the U.S. AirForce, began in January 1967 [1]. The
objective wasthe development of a ground-based radar that
coulddetect intruders moving into a small encampment.This
ground-based system was named the Camp Sen-tinel Radar. Deployed in
Vietnam after an eighteen-month crash development program, it
protectedAmerican troops throughout the rest of the war.
At the inception of the Camp Sentinel Radar pro-gram, there was
little information on which to basethe radar design. The technical
literature providedminimal data on electromagnetic attenuation
intropical foliage, but it did suggest that the best operat-ing
frequencies were between 20 and 500 MHz. Afrequency (435 MHz) near
the upper end was chosenin order to localize any detections to a
small azi-muthal region with an antenna small enough for tac-tical
deployment.
Three critical questions had to be answered aboutthe propagation
of radar signals within relativelydense foliage: (1) How much does
the moving foliagespread the frequency spectrum of a signal
reflectedfrom a target? (2) What is the frequency spectrum
ofclutter signals reflected from windblown foliage? (3)What is the
effect of multipath propagation on rangeand azimuth resolution and
on subclutter visibility?
Lincoln Laboratory built two radar systems to an-swer these
questions. The initial system, called theCamp Sentinel Radar-I,
took measurements at a sitelocal to Lincoln Laboratory. This system
worked well,and it was used in demonstrations to military
observ-ers in the fall of 1967. A second version was mountedin a
van and sent to Bisley, Puerto Rico, where foliageclosely simulated
the conditions in Vietnam. By Janu-ary 1968, enough measurement
data had been accu-mulated to go ahead with a more advanced
radar,called the Camp Sentinel Radar-II.
The design of the Camp Sentinel Radar-II incor-porated unique
and innovative concepts. The an-
tenna was mounted high above the ground on a rap-idly deployable
tower so the electromagnetic wavescould reach a target by
propagating over the tops ofthe trees and then be diffracted to the
ground, ratherthan by propagating directly through the foliage.
Anelectronically scanned cylindrical array sequentiallystepped the
antenna beam through thirty-two posi-tions in azimuth to cover
360°. The transmit/receivebeams were stepped so rapidly in azimuth
that thesignal processing functioned as if there had beenthirty-two
individual radars, each with stationary azi-muth coverage and all
operating simultaneously. Theprocessors were able to execute the
critical algorithmsneeded to detect small, slow-moving human
infiltra-tors from the high-level clutter background createdby the
windblown tropical foliage.
The radar control was designed to allow an opera-tor to
construct two intrusion fences. These fencescould be made irregular
in shape, to match them tothe desired defense perimeter. The
operator did notneed to monitor the radar unless an alarm sounded.
Ifa detection occurred, the operator simply checked thedisplay to
see which range/azimuth sector containedthe intruder and whether
the target was incoming oroutgoing. After a short local testing
period, the firstLaboratory-built Camp Sentinel Radar-II was
FIGURE 1. The Camp Sentinel Radar-II was first installedand
operated in Lai Khe, Vietnam, in 1968. The system wassituated on
the perimeter of the U.S. Army camp, with theradar antenna mounted
on top of the tower. The remainderof the radar system was located
in a bunker behind the hill inthis picture, where the machine-gun
tower was located.
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 343
shipped to Vietnam in August 1968. Laboratory em-ployees Leonard
Bowles and David Rogers spent twomonths in Vietnam, introducing the
radar to theThird Brigade, 1st Infantry Division, and
instructingArmy personnel in its operation and maintenance.The
radar, shown in Figure 1, received immediate ac-ceptance, and the
Army used it until the end of thewar.
The U.S. Army’s Harry Diamond Laboratory car-ried out additional
development work. The improvedversion, called the Camp Sentinel
Radar-III, includeda more powerful transmitter to increase the
detectionrange and to provide an additional number of
displayoptions. Six of these radars were manufactured andsent to
Vietnam, where they remained until U.S.combat troops were
withdrawn.
Geodar: Ground-Penetrating Radar (1966–1967)
American forces operating in Vietnam needed a sen-sor system
that could detect tunnels. In 1966, afterdiscussions with other
laboratories working on thisproblem, Robert Lerner of Lincoln
Laboratory con-cluded that a ground-penetrating radar offered
possi-bilities, and the Geodar (ground echo detection andranging)
program was initiated to investigate the con-cept under DARPA
sponsorship. The distinguishingfeature of the Geodar concept was
that electromag-netic energy was radiated directly into the
ground.
Because the soil-penetration properties of radarwere not known
with any precision, it was decided touse a wide band of
frequencies, from 50 to 150 MHz,within the generally applicable
frequency range. A flatrectangular-shaped antenna of
transmission-line de-sign, operating close to the ground surface,
radiatedelectromagnetic energy in compact packets of 3-to-5-nsec
duration. The antenna, shown in Figure 2, withan effective area of
about 3/4 of a square meter, wasdrawn over the ground on a Teflon
sled structure.
A first experimental system was quickly assembledin 1966, and
proof-of-concept tests were made at asimple tunnel test range at
Lincoln Laboratory. Thetests proved that tunnel-like voids in the
groundcould be detected. A formal program was then estab-lished to
develop a demonstration system for a fieldtest. The first system,
Geodar Mark I, was completedin March 1967, and an improved version,
GeodarMark II, was completed a few months later [2].
The Geodar systems were tested on tunnels at FortBelvoir,
Virginia, and at Raleigh, North Carolina, aswell as on voids
implanted in a second Laboratory testrange constructed at Millstone
Hill in Westford, Mas-sachusetts. The test results indicated that
the Geodarsystems could locate tunnels of two to three feet in
di-ameter at depths of up to approximately twenty feetin most
alluvial, glacial, and loessial soils.
Several sets of Geodar Mark II were fabricated bySylvania West,
and Lerner went with them to Viet-nam. Demonstrations there
corroborated the earliertest data and predictions. For a time, the
Geodar sys-tem was deployed for perimeter tunnel surveillancearound
a U.S. Army Headquarters installation.
Hostile Weapons Location System (1974–1981)
The joint DARPA–U.S. Army Hostile Weapons Lo-cation System
(HOWLS) program, which began in1974, focused on the development of
techniques tolocate and classify stationary indirect-fire weapons
inbackground clutter. In December 1974, the GeneralElectric Company
was selected to build the HOWLSairborne radar. The HOWLS system
consisted of aKu-band (16 GHz) array radar mounted in a
twin-en-gine aircraft, a ground-station recording system, andan
off-line processing system in a van [3]. The azi-muth resolution
(real-beam) was 0.5° and the range
FIGURE 2. Field operation of the Geodar ground-penetrat-ing
radar system. Initial testing of this system consisted ofdragging
the antenna over a known test area to detect tun-nel-like voids
under the surface. The antenna was con-nected to the radar
equipment in the jeep by a cable.
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
344 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
resolution was ten meters, which was a high-resolu-tion system
at that time. The program addressed twomajor issues: solve the
automatic-target-detectionproblem, and implement this solution on a
miniatureunmanned air vehicle (UAV) radar.
Figure 3 shows the airborne and ground-stationcomponents of the
HOWLS experimental radar sys-tem. Figure 3(a) shows the HOWLS radar
antennamounted under the fuselage of a twin-engine PiperNavajo
aircraft. The radar was a Ku-band system witha 500-MHz
radio-frequency (RF) bandwidth. Themaximum coherent-pulse bandwidth
was 18 MHz(corresponding to a resolution of ten meters), and
thecenter frequency could be stepped across the RFbandwidth in
sixty-four steps. The system was used todevelop
stationary-ground-target detection tech-niques with a spatial
resolution of 10 × 10 m and asingle polarization. Figure 3(b) shows
the interior ofthe ground-station van with the signal-processing
and recording system. The HOWLS system provided the
information needed to design a lightweight and low-cost radar
appropriate for mini-UAV applications.
Figure 4 shows a HOWLS ground-radar map ofStockbridge, New York,
at 10 × 10-m resolution withtarget detections overlaid. The white
squares form acalibration array, while the red squares are
detectionsof targets such as eight-inch guns and armored ve-hicles.
Speckle in the image was reduced by non-coherently averaging over
sixteen independent fre-quencies. Targets in relatively open areas
could bedetected, but the false-alarm rate was high from natu-ral
and man-made clutter. Resolution was still inad-equate for
acceptable classification of stationary tacti-cal targets.
One major accomplishment of HOWLS was thedevelopment of a new
theory of target detection thatmore accurately predicted the
detection performanceof moderate-resolution radar systems [4].
Earliertheories that used noise models to represent groundclutter
gave very optimistic detection results, whichwere in error by one
to two orders of magnitude. Thenew theory correctly took clutter
inhomogeneity intoaccount, leading to more accurate predictions of
per-formance. Figure 5 plots the probability of detectionof armored
targets against the number of false alarmsexpected per square
kilometer. The dashed black lineis the fit to the experimental
data; the average target-
FIGURE 4. A HOWLS radar map of the ground atStockbridge, New
York. The white squares are returns froma calibration array, while
the red squares are detections oftargets such as eight-inch guns
and armored vehicles.
FIGURE 3. (a) The airborne component of the DARPA–U.S.Army
Hostile Weapons Location System (HOWLS) in-stalled in a twin-engine
Piper Navajo aircraft with the arrayantenna mounted below the
fuselage. (b) Data from thissensor were linked to a ground-station
van for signal pro-cessing, recording, and display.
(a)
(b)
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 345
with good resolution like the HOWLS radar wouldbe inadequate for
target detection because the sensorwould be swamped with false
alarms if it were set toachieve realistic target-detection
probabilities. Toachieve the higher detection probabilities with a
sig-nificantly lower false-alarm density required muchbetter
resolution, such as that achievable with a syn-thetic-aperture
radar (SAR) and a polarimetric-mea-surement capability to further
distinguish targetsfrom natural and cultural clutter. The
performancethat can be realized with such a system is
discussedlater in this article in the section entitled
“Stationary-Target Detection Employing High Resolution
andPolarization.”
Netted Radar Program (1976–1981)
The Netted Radar Program began in 1976 to addressthe
moving-target detection problem and to demon-strate that an
operational system of distributed radarsconnected by a
multiple-user network would providefor the first time a
comprehensive view of the battle-field. In addition, the program
showed that the U.S.Army’s AN/PPS-5 ground radars could be vastly
im-proved by adding modern digital signal-processingtechniques
originally developed by Lincoln Labora-tory for use in air-traffic
control.
Figure 6 illustrates the Netted Radar System asdemonstrated at
Fort Sill, Oklahoma. The networkutilized an airborne radar (a
modified version of theHOWLS radar called the Advanced Airborne
Radarwith moving-target detection and tracking capability[5]), two
modified AN/PPS-5 ground-based radars,and a U.S. Army AN/TPQ-36
ground-based artil-lery-locating radar. The AN/PPS-5 radars
wereequipped with modern signal-processing capabilitiesimplemented
by Lincoln Laboratory, and were re-named AN/TPS-5X radars [6].
Figure 7(a) shows anAN/TPS-5X antenna looking over Fort Sill,
Okla-homa, from nearby Mount Scott. The AN/TPS-5Xradar could detect
and track tank-sized targets out toa range of about twenty
kilometers. In addition to de-tecting moving ground vehicles, the
AN/TPS-5X ra-dar also detected low-flying aircraft, rotating
anten-nas, moving personnel, and artillery shell bursts. Itcould
also estimate the azimuth positions of noisejammers.
FIGURE 5. Detection performance of HOWLS against sta-tionary
targets in clutter. The assumption of a homogeneousground-clutter
model predicts very optimistic results, asshown in the upper black
curve for a target-to-clutter ratio(T/C) of 6 dB. The HOWLS radar
data from experiments inStockbridge, New York, shown as red points,
show the prob-ability of detection of armored targets versus the
number offalse alarms per square kilometer. The curve-fit
HOWLSdata, shown as a black dashed line, indicate a much less
op-timistic performance for this medium-resolution radar thanthe
homogeneous-clutter model, an error of about two or-ders of
magnitude. A new theory developed by Leslie M.Novak at Lincoln
Laboratory assumes a nonhomogeneous-clutter model, and predictions
from this theory bracket theexperimental results quite well, as
shown by the blue dashedlines for T/C = 5 dB and T/C = 8 dB
[4].
1.0
0.8
0.6
0.4
0.2
01 10 100 1000
False alarms/km2
Pro
babi
lity
of d
ectio
n
(T/C = 6 dB)
Experimental data (T/C ~6 dB)
(T/C = 5 dB)
(T/C = 8 dB)
Homogeneous-clutter calculationNonhomogeneous-clutter
calculation
to-clutter ratio (T/C ) was 6 dB. The theoreticalcurves for T/C
= 5 dB and T/C = 8 dB nicely bracketthe curve for the experimental
data. The homoge-neous-clutter model (the top curve in Figure 5)
pre-dicts 2 false alarms/km2 for a detection probability of0.8 and
a T/C = 6 dB, while the experimental-datacurve shows 200 false
alarms/km2 for the same condi-tions—an error of two orders of
magnitude. TheHOWLS program demonstrated the potential ofboth
automatic-target-detection techniques for sta-tionary targets and
the utility of mini-UAV radar sys-tems with their excellent terrain
visibility.
The HOWLS program was a pioneering
wide-areabattlefield-surveillance effort. It proved that a
sensor
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
346 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
All radars were netted together through a TargetIntegration
Center, which in turn was connected tothe Army’s Tactical Artillery
Fire Direction System(TACFIRE) computer center. The system
providedexceptional coverage of the battlefield and greatly
re-duced the delay between target detection and artilleryfire on
the target. This demonstration clearly showedthe Army the
advantages of tracking and reportingmoving targets with an
automatic real-time system,compared to the Army’s established
“man-in-the-loop” methods.
Figure 7(b) illustrates the operator display for theNetted Radar
Demonstration. Each tracked target isautomatically assigned a
target number that can beaccessed to give the target’s location and
velocity. Forinstance, target 17 is a ground object moving at
12m/sec, while target 16 is a helicopter in the air. Thesystem also
provided jammer location and artillery-fire registration by using
range and azimuth triangu-lation techniques.
The Fort Sill demonstration clearly showed the ad-vantages of an
airborne radar for terrain visibility.
Even in the moderately level terrain of Oklahoma,terrain masking
was a serious problem for ground-based sensors, while virtually all
targets were visible tothe airborne radar. The airborne radar
component ofthe Fort Sill demonstration pointed to the value ofUAV
radar with moving-target detection and track-ing capability.
In addition to rapid target acquisition, the advan-tages of
combining target information with rapid dis-semination to artillery
units were demonstrated inreal time. As an example of the speed and
accuracy ofthe networked system, a tank, fortified to withstandhits
with inert artillery rounds, was tracked by the ra-dar network as
it moved along a roadway in the EastRange target area at Fort Sill.
A firing time was com-puted at the Target Integration Center on the
basis of(1) the predicted arrival time of the tank at the
prede-termined aim point on the road, and (2) the esti-mated flight
time of the projectile from gun to aimpoint. Because the artillery
had been zeroed in on theaim point, and because the predicted
projectile timeof arrival derived from the real-time
measurements
FIGURE 6. Elements of the Netted Radar Program demonstration of
moving-target detection.Data from the single airborne radar and
three ground-based radars were sent over narrowbandVHF links to the
Target Integration Center, where they were integrated with the
Army’s TacticalArtillery Fire Direction System (TACFIRE) computer
center. Fire missions then brought artil-lery fire against
designated moving targets on roads in real time.
Airborne radar
AN/TPS-5X radarsupport van
AN/TPQ-36radar
TACFIREcomputer center
AN/TPS-5X radarsupport van
TargetIntegration
Center Airborne radarsupport van
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 347
was very accurate, the inert round hit the tank anddamaged one
of the tank treads. This convincing re-sult demonstrated the
improved performance of theAN/TPS-5X radars, the automatic
target-acquisitionand tracking functions, the feasibility of radar
net-ting, and the potential to provide for more
accurateartillery-fire adjustment.
The basic objective of the Netted Radar Programwas to develop
the technology for the netting of
battlefield radars. The AN/TPS-5X radars had severallimitations:
they could not serve multiple users in adynamic situation, they
were easily detected by theirscan motion, and the VHF radio system
was not jam-resistant enough for a tactical configuration.
For these reasons, Lincoln Laboratory developedthe Advanced
Ground Surveillance Radar (AGSR),which was incorporated as a
parallel element of theNetted Radar Program. The AGSR was a
moving-tar-get-indicator (MTI) radar. It featured a C-band
cylin-drical array antenna, which was electronicallysteerable in
azimuth over 360° and capable of simul-taneous multimode radar
operation together with anintegral data link. [7] The fully
coherent MTI AGSRwas demonstrated at Fort Sill from November
1980through January 1981, where it automatically trackedground
vehicles, walking troops, and helicopters. Italso detected and
accurately located artillery shellbursts.
The achievement of the Netted Radar Programwas the completely
automated fusion of a number ofsurveillance radars in U.S. Army
exercises under vari-ous field conditions. The system developed
under thisprogram was conceptually similar to the Laboratory’sSAGE
(Semi-Automatic Ground Environment) air-defense system developed in
the 1950s [8]. The suc-cess of this program set the stage for the
netting ofother operational military surveillance radars.
Unmanned-Air-Vehicle RadarProgram (1982–1991)
In 1982, under DARPA and U.S. Army sponsorship,Lincoln
Laboratory began a UAV-radar developmentprogram to detect and
classify moving targets with alow-power, lightweight radar designed
for a UAVplatform. The system provided either a specific angu-lar
sector or full 360° surveillance of moving groundvehicles and
low-flying helicopters, as illustrated inFigure 8. MTI data were
first processed onboard theUAV. Low-bandwidth data were then linked
to aground station where precision track, location,
andclassification of moving targets were performed andthe results
were displayed [9].
Lincoln Laboratory researchers designed the UAVradar, using
commercially available components forthe RF parts of the radar such
as the transmitter, re-
FIGURE 7. (a) The radio-frequency (RF) portion of a U.S.Army
AN/TPS-5X radar looking over Fort Sill, Oklahoma,from Mount Scott.
A nearby radar support van housed thesignal-processing and
data-processing equipment. (b) Anillustration of the operator
display used in the Netted RadarDemonstration. The Fort Sill
demonstration clearly showedthe advantages of an airborne radar for
significantly reduc-ing terrain masking.
5
4
7
8
6
3
2
1
0
0 1 2 3 4 5
2421 17 23
22
26
16
Current estimatedtruck location
=
Recent track history=
10
6 7 8
Kilometers
Kilo
met
ers
(b)
(a)
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
348 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
ceiver, and antenna. One important issue was to pro-vide onboard
MTI processing and to provide a low-bandwidth communication link to
send the target re-ports to a ground station for classification
andreal-time display. Since a suitable commercial proces-sor was
not available at that time, the Laboratory de-veloped a
state-of-the-art processor together with thenecessary
signal-processing and data-processing algo-rithms. Figure 9 shows
the equipment as mounted inan Amber UAV fuselage. Figure 10 shows
how theentire system was form-fitted in the UAV fuselageand
captive-carried on a Twin Otter aircraft for test-ing and
evaluation. The radar could be operated fromthe ground station or
independently by an operator
in the aircraft. The complex radar data were recordedon a
13-MB/sec recorder, so missions could be flownbeyond the reach of
the ground station.
In the spring of 1990, a field demonstration wasconducted at
Fort Sill and evaluated by the U.S.Army’s Intelligence School. As
an example of real-time wide-area surveillance, Figure 11 shows the
de-tection of virtually all of the moving targets found inless than
three minutes within a 900-km2 area sur-rounding Fort Sill (above
center in the figure) and thetown of Lawton, Oklahoma (just below
center). Thedensity of moving targets is clearly visible from
thenumber of MTI detections overlaid on the digitizedroad map.
Notice the absence of detections or false
FIGURE 8. The unmanned-air-vehicle (UAV) surveillance and
tracking radar concept. (a) The remotely con-trolled UAV carries a
small coherent moving-target detection radar that provides
surveillance of a large areaof the surrounding terrain and sends
moving-target reports, generated by an onboard processor, via data
linkto a ground station for display and target tracking. UAV
position estimates are provided by a Global Position-ing System
(GPS) receiver and are used to update an inertial-navigation system
that provides filtered UAV po-sition (two dimensions), altitude,
and aircraft attitude information. (b) With high look-down angles,
the UAVradar has an excellent view of the surrounding terrain for
detection and tracking. With its small visible signa-ture and low
radar cross section, the UAV is difficult to locate and destroy
with surface-fired weapons.
Radar at 3000-m altitude
5 km
(b)
(a)
37°
3° azimuth
11°
15 km
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 349
FIGURE 11. UAV-radar wide-area MTI surveillance at FortSill,
Oklahoma. An activity map of moving vehicles showsthe vehicular
traffic near Fort Sill and the town of Lawton,Oklahoma. This map
represents moving-target detectionsobtained from several 360° scans
of the antenna.
FIGURE 9. The moving-target-indicator (MTI) radar mounted in the
UAV fuselage. The open panels show thelocation of the radar
components within the fuselage. The UAV, with a twenty-foot wing
span, was developedby Leading Systems, Irvine, California, under
DARPA sponsorship, and was identified by the name “Amber.”
FIGURE 10. The UAV-radar captive-carry flight-test
configu-ration. The UAV fuselage was attached, without wings,
tail,or engine, to a Twin Otter aircraft for testing. All the UAV
ra-dar equipment was contained within the UAV’s fuselage.The Twin
Otter aircraft flies at a speed comparable to that ofthe UAV.
alarms in the (restricted access) artillery ranges to
thenortheast and northwest. The accuracy of targetplacement can be
qualitatively measured by howclosely detections match the
underlying road grid.
The radar was also required to supply moving-tar-get reports and
tracks against a background map ofthe Fort Sill area. During a
blind test, the radar opera-tor was asked to determine the number
of vehicles,the speed, and the mix of wheeled and tracked ve-hicles
in the convoys sent out by the evaluation team.The radar
successfully reported location, speed, andcomposition of the
convoys out to ranges of sixteen
kilometers. The field demonstrations were conductedjointly with
the Army Research Laboratory. The com-bined efforts and the
demonstration results con-vinced Army users of the value of such an
MTI radaron a UAV. The Joint UAV Program Office is nowconsidering
such applications.
Kilometers
Kilo
met
ers
30201000
10
20
30
Exciter/receiver
Data link andGPS receiver
Signal processor
Inertial navigation system
Antenna subsystem
Traveling-wave-tube amplifier
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
350 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
Lincoln Laboratory has made many significantcontributions to the
improvement of airborne MTIsystems. Much of the pioneering work in
the area ofsurface surveillance has contributed directly to the
de-velopment of national assets such as the U.S. AirForce Joint
Surveillance Target Attack Radar System(Joint STARS) airborne
surveillance system. Thiswork included the development of
improvements toMTI target-detection techniques, SAR imaging,
andfixed-target detection, as discussed in this article. Amajor
contribution in the area of airborne MTI clut-ter cancellation and
target angle estimation was madewith the concept of the
displaced-phase-center an-tenna (DPCA), which introduced the
multiple-phase-center array concept. The details of DPCA arecovered
in the article entitled “Displaced-Phase-Cen-ter Antenna
Technique,” by Charles Edward Mueheand Melvin Labitt, in this
issue.
Joint STARS is a long-range
surface-surveillancemultiple-aperture array radar carried by U.S.
AirForce E-8C aircraft. The wide-area surveillance andmoving-target
indicator (WAS/MTI) are the radar’sfundamental operating modes.
WAS/MTI is de-signed to detect, locate, and identify slow-moving
tar-gets. Synthetic-aperture radar/fixed-target indicator(SAR/FTI)
provides high-resolution SAR imaging forstationary-target detection
and identification.
Stationary-Target Detection EmployingHigh Resolution and
Polarization (1982–1996)
In parallel with the UAV MTI work, some of theLaboratory’s
effort in tactical technology shifted todeveloping more capable
techniques for stationary-target detection and understanding the
performancelimits of these techniques. The Laboratory began
theAdvanced Detection Technology (ADT) program in1982 under DARPA
sponsorship in response to aDepartment of Defense need to examine
the poten-tial of MTI radar for use in “smart weapons.” A
verycapable instrumentation radar, called the AdvancedDetection
Technology Sensor (ADTS), was built forLincoln Laboratory by
Goodyear Aerospace. TheLaboratory used this dual-polarized sensor
to capturethe full dimensionality of the radar signal. The
radar,mounted in a Gulfstream II aircraft as shown in Fig-ure 12,
provided well-calibrated, high-resolution,
FIGURE 12. The Lincoln Laboratory airborne Advanced De-tection
Technology Sensor (ADTS) had X-band, Ku-band,and Ka-band
(monostatic and bistatic) synthetic-apertureradar (SAR) modes.
Either (a) the dual-band (X and Ku) an-tenna (reflector type) or
(b) the Ka-band horn antenna ismounted in (c) the radome below the
Gulfstream II aircraftfuselage, depending on the data-collection
requirements.
fully polarimetric, real or synthetic-aperture data.Originally
built for Ka-band (33 GHz) operation[10], the radar was later
modified to include X andKu-bands together with a bistatic
capability as well.
This radar and the associated research programshave played a
vital role in the national effort to de-velop automatic target
recognition (ATR) of station-ary ground targets. The airborne data
acquired by theradar in Figure 12 have been a principal source for
de-velopment in the ATR research community. For ex-ample, the
high-resolution SAR map of the site inStockbridge, New York,
displayed in Figure 13(a)shows targets in typical background
clutter. Figure13(b) shows the declared detections that result
fromprocessing the data represented in Figure 13(a). Fig-ure 13(c)
shows the results of an end-to-end perfor-mance study using a
baseline SAR ATR algorithmsuite. These results have formed the
fundamental un-derstanding of the value of SAR resolution and
mul-tiple polarizations in ATR.
Figure 13(c) summarizes the fundamental trade-offs between
false-alarm density versus probability of
(a) (b)
(c)
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 351
FIGURE 13. (a) The ADTS high-resolution SAR map of
theStockbridge, New York, site covered approximately the
samegeographical area of natural and cultural clutter as shown
inthe low-resolution map in Figure 4. In this high-resolutionSAR
map, light pixels indicate strong radar reflections. Lo-cated on
the right side of the image is a region containingvarious tactical
targets. On the left side of the image is alarge maintenance area
containing buildings, vehicles, air-craft, and other man-made
objects. (b) Declared detectionsare indicated as white markers on
the ADTS SAR map ofthe Stockbridge site. Note the relative absence
of false de-tections in the natural-clutter areas. (c) SAR
automatic-tar-get-recognition (ATR) detection performance results
versuspolarization and resolution for the Stockbridge site. HH
rep-resents horizontal polarization (on both transmit and
re-ceive), and PWF represents the polarimetric whitening
filter.
detection for single and multiple polarizations (opti-mally
combined) and for various resolutions [11, 12].Algorithms developed
under these programs wereused in the DARPA Semi-Automated IMINT
(ImageIntelligence) Program, or SAIP.
Over four hundred data-collection missions were
flown with the ADTS aircraft, satisfying a variety ofmission
objectives relating to (1) detection and classi-fication of
stationary tactical targets, (2) internal sea-wave detection for
antisubmarine warfare, (3) bistaticphenomenology for missile-seeker
applications, and(4) SAR for moving-target imaging. Following an
ac-tive and productive lifetime, operations with theADTS sensor
ended in late 1998.
Synthetic-Aperture Foliage-PenetrationRadars (1987–1996)
Synthetic-aperture foliage-penetration (FOPEN) ra-dars have been
developed because of the need to findstationary tactical targets
located in deep foliage andnatural camouflage, which hide these
targets fromconventional microwave surveillance sensors. Duringthe
Vietnam era, as noted earlier, military personnelsuccessfully used
the Camp Sentinel ground-based ra-dar to detect enemy soldiers and
vehicles moving infoliage. To locate stationary targets,
researchers alsoneeded to develop a low-frequency
foliage-penetra-tion SAR imaging capability with sufficient
resolu-tion to differentiate targets from clutter. Unfortu-nately,
no reliable target-recognition algorithms wereavailable at that
time to reduce false alarms. Many ofthe currently successful ATR
and cueing techniquesfor detecting moving and stationary targets
have beendeveloped to be used only for targets in the open.
Since the late 1980s, Lincoln Laboratory, underDARPA and U.S.
Air Force sponsorship, has plannedand conducted a number of
experiments and data-collection programs utilizing a variety of
industry-built sensors to evaluate the use of low-frequency ra-dar
to detect and identify tactical targets hidden byfoliage. The
results from these efforts have been usedto develop the current
automatic target-detection andcueing algorithms that operate on UHF
and VHFSAR data. When we use a SAR system to detect ob-jects hidden
or obscured by foliage, detection is de-graded in three ways for
the higher microwave fre-quencies (>100 MHz). First, the foliage
contributesto the clutter return. Second, the foliage
attenuatessignal propagation through it. And third, the
movingfoliage induces fluctuations in the amplitude andphase of the
radar signal, which distort the SAR im-age of the target. These
fluctuations affect the image-
1.0
0.8
0.6
0.4
0.2
0.00.01 0.1 1 10 100 1000
False alarms per km2
Pro
babi
lity
of d
etec
tion
PWF1 ft × 1 ft
PWF1 m × 1 m
HH1 ft × 1 ft
HH1 m × 1 m
End-to-end performanceof ATR algorithm suitetactical targets
(netted)
Stockbridge clutter(56 km2 total)
(c)
(a)
(b)
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
352 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
focusing quality and the detection performance ofSAR for targets
hidden by foliage.
To obtain a better understanding of the foliage ef-fects,
researchers needed an accurate quantitative as-sessment of these
issues. In 1990, Lincoln Laboratoryconducted a definitive
experiment to measure foliageattenuation and backscatter of heavily
forested areas.This experiment was conducted with the
NASA/JetPropulsion Laboratory AIRSAR aircraft, which hasUHF,
L-band, and C-band SAR radars. Figure 14shows this aircraft
equipped with these systems.
The Laboratory conducted tests by overflying aforest site in
northern Maine. The site had been care-fully mapped and implemented
with a variety of cali-bration reflectors and devices. Phase and
amplitudedata from the experiment were coherently integratedto
create synthetic-aperture azimuthal patterns thatwould result when
imaging a point target obscured bythe foliage. The effects of
synthetic-aperture length,frequency, and polarization on the
attenuation andazimuthal synthetic beam pattern were
investigated.Measurements also showed that less than a
one-meterresolution for foliage penetration could be achieved atUHF
frequencies. The results, shown in Figure 15,demonstrate the
well-defined synthetic-aperture azi-muthal patterns that can be
generated even down toresolutions as small as 0.6 meters. This
result was cer-tainly a surprise to the conventional wisdom of
thetime, and it launched a major Department of Defenseeffort in
FOPEN radar technology.
In the past decade, this FOPEN work has been ex-tended to
develop a phenomenological understand-ing of foliage penetration
and to develop advanced
FIGURE 15. Foliage-penetration (FOPEN) results demon-strate that
SAR azimuthal patterns can be generated with in-creasing resolution
as small as 0.6 m, which indicates theability of FOPEN systems to
form accurate SAR beamsthrough foliage.
0
–20
–10
–30
–40
–50
–60–20 –15 –10 –5 0 5 10 15 20
Cross range (m)
Rel
ativ
e po
wer
(dB
)
4 m2 m1 m0.6 m
Resolution
automatic-target-detection and recognition tech-niques.
Additional foliage-penetration measurementswere made in 1993 in
tropical rain forest and north-ern U.S. forest environments by
using the Swedish3-m-resolution Coherent All Radio Band
Sensor(CARABAS) and the Stanford Research Instituteultrawideband
(UWB) SAR sensor collecting hori-zontal-polarization VHF and UHF
data. The datacollected in these experiments were processed
intocalibrated SAR imagery, and foliage-induced attenua-tion for
all frequency bands was calculated by com-parison of echoes from
test reflectors in foliage andthose in the open [13–15].
In support of the current DARPA-sponsored foli-age-penetration
SAR project, Lincoln Laboratory iscurrently developing
automatic-target-detection andcueing algorithms for VHF and UHF
radar. With thedetection of targets now possible with our existing
fo-liage-penetration capability, the remaining issue be-comes the
identification of threatening targets fromall of the many
detections that are reported. Twofalse-alarm-mitigation
techniques—change detectionand group detection—have been used to
reduce thesefalse alarms substantially.
Wideband FOPEN systems are needed to detecttargets reliably in
the presence of foliage, radio inter-ference, and cultural clutter.
This requirement has
FIGURE 14. The NASA/Jet Propulsion Laboratory AIRSARaircraft
containing the UHF, L-band, and C-band SAR ra-dars. This aircraft
was used to measure and characterize ra-dar attenuation and
backscatter caused by foliage in heavilyforested areas.
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
VOLUME 12, NUMBER 2, 2000 LINCOLN LABORATORY JOURNAL 353
motivated the development of a UWB SAR operatingover 215 to 730
MHz. Its resolution is 0.3 m in rangeand 0.6 m in cross-range; and
it has a full range ofpolarizations. The radar was built by the
Environ-mental Research Institute of Michigan and installedon a
U.S. Navy P-3 aircraft controlled by the NavalAir Warfare Center.
This test-bed sensor, funded bythe Air Force Wright Laboratory, was
flown in 1995to collect data on tactical targets in geographically
di-verse foliage over some 1500 km2 of foliated terrain.Lincoln
Laboratory has analyzed these data to deter-mine the system
requirements for an operational sys-tem and to identify suitable
concepts to be employedwith a tactical platform. Preliminary
results from theUWB data suggest that a resolution finer than
onemeter is required to reduce false alarms from trees andto
provide a sufficient number of independent pixelsfor recognition of
potential targets in a radar image.At UHF frequencies, this
requirement implies weneed SAR azimuth integration angles greater
than 35°with correspondingly long integration times.
LincolnLaboratory has also designed a new coherent classifierto use
the complex polarimetric UHF data. The clas-sifier performs well on
targets in the open, and it cor-rectly classifies a significant
portion of targets hiddenunder foliage.
Recently, many researchers have been interested inthe use of an
airborne SAR for detection of under-ground targets, large and
small, such as mines andtrucks hidden in underground bunkers. In
support ofthis interest, a ground-penetration experiment
wasconducted in 1993 near Yuma, Arizona. Again, a vari-ety of
radars were used, covering the RF band from20 to 500 MHz. Data from
this test helped research-ers to develop a phenomenological
understanding ofsoil-penetration losses and clutter backscatter,
and toinvestigate the signatures of buried targets [16].
Summary
Lincoln Laboratory has had a major influence overthe past thirty
years on the science and technology ofbattlefield surveillance for
stationary and moving tar-gets. Major developments and tests of
airborne andground sensors have been accomplished.
In all these efforts the goals have been to (1) under-stand the
fundamental radar and processing issues in-
volved, (2) determine the theoretical performancelimits based on
scientific principles and state-of-the-art technology, and (3)
confirm those limits with fielddemonstrations. Many problems still
remain inbattlefield surveillance, and the ongoing program atthe
Laboratory continues to address these challenges.
R E F E R E N C E S1. E.C. Freeman, ed., MIT Lincoln Laboratory:
Technology in the
National Interest (Lincoln Laboratory, Lexington, Mass.,1995),
pp. 153–155.
2. MIT Lincoln Laboratory: Technology in the National
Interest,p. 156.
3. V.L. Lynn, “HOWLS Radar Development,” in MillimeterWave
Radars, S.L. Johnston, ed. (Artech House, Dedham,Mass., 1980), pp.
621–634.
4. L.M. Novak and F.W. Vote, “Millimeter Airborne Radar Tar-get
Detection and Selection Techniques,” IEEE NAECON-79Conf. Record 2,
Dayton, Ohio, 15–17 May 1979, pp. 807–817.
5. MIT Lincoln Laboratory: Technology in the National
Interest,p. 160.
6. M.I. Mirkin, C.E. Schwartz, and S. Spoerri,
“AutomatedTracking with Netted Ground Surveillance Radars,”
IEEE1980 Int. Radar Conf., Arlington, Va., 28–30 Apr. 1980,
pp.371–379.
7. J.-C. Sureau, “Applications of Cylindrical Arrays to
Surveil-lance Radars,” Int. Conf. on Radar, Paris, 4–8 Dec. 1978,
pp.337–343.
8. MIT Lincoln Laboratory: Technology in the National
Interest,pp. 15–33.
9. MIT Lincoln Laboratory: Technology in the National
Interest,pp. 158–159.
10. J.C. Henry, “The Lincoln Laboratory 33 GHz Airborne
Pola-rimetric SAR Imaging System,” IEEE National TelesystemsConf.
Proc., Atlanta, 26–27 Mar. 1991, pp. 353–358.
11. L.M. Novak, M.C. Burl, and W.W. Irving, “Optimal
Polari-metric Processing for Enhanced Target Detection,” IEEETrans.
Aerosp. Electron. Syst. 29 (1), 1993, pp. 234–244.
12. L.M. Novak, S.D. Halversen, G.J. Owirka, and M. Hiett,
“Ef-fects of Polarization and Resolution on the Performance of aSAR
Automatic Target Recognition System,” Linc. Lab. J. 8(1), 1995, pp.
49–68.
13. J.G. Fleischman, S. Ayasli, E.M. Adams, and D.R.
Gosslen,“The July 1990 Foliage Penetration Experiment, Part I:
FoliageAttenuation and Backscatter Analysis of SAR Imagery,”
IEEETrans. Aerosp. Electron. Syst. 32 (1), 1996, pp. 135–144.
14. M.F. Toups, S. Ayasli, and J.G. Fleischman, “The July
1990Foliage Penetration Experiment, Part II: Analysis of
Foliage-Induced Synthetic Pattern Distortions,” IEEE Trans.
Aerosp.Electron. Syst. 32 (1), 1996, pp. 145–155.
15. B.T. Binder, M.F. Toups, S. Ayasli, and E.M. Adams,
“SARFoliage Penetration Phenomenology of Tropical Rain Forestand
Northern U.S. Forest,” IEEE Int. Radar Conf., Alexandria,Va., 8–11
May 1995, pp. 158–163.
16. M.I. Mirkin, T.O. Grosch, T.J. Murphy, S. Ayasli, H.
Hellsten,R. Vickers, and J.M. Ralston, “Results of the June 1993
YumaGround Penetration Experiment,” SPIE 2217, 1994, pp. 4–15.
-
• BRYANT, MORSE, NOVAK, AND HENRYTactical Radars for Ground
Surveillance
354 LINCOLN LABORATORY JOURNAL VOLUME 12, NUMBER 2, 2000
. was leader of the SurveillanceSystems group until he
retiredfrom Lincoln Laboratory in1996. He is now working forthe
Laboratory as a consultant.His research interests includeradars and
signal processing.Before joining the Laboratoryin 1966, he worked
for VarianAssociates. He received a B.S.degree in physics from
North-eastern University and anM.S. degree in physics
fromRensselaer PolytechnicInstitute.
. is a staff member in the Sur-veillance Systems group. Hejoined
Lincoln Laboratory in1968. His background is inradar systems and
signal pro-cessing. His current interest isin airborne
synthetic-apertureradar for surface surveillanceand target
classification.He received a B.S.E.E. degreein 1966 and an
M.S.E.E.degree in 1968 from theUniversity of Maine. He is asenior
member of the IEEE,Eta Kappa Nu, Tau Beta Phi,Phi Kappa Phi, and
Sigma Xi.
. is a senior staff member in theSurveillance Systems group.He
received a B.S.E.E. degreefrom Fairleigh DickinsonUniversity in
1961, anM.S.E.E. degree from theUniversity of Southern Cali-fornia
in 1963, and a Ph.D.degree in electrical engineeringfrom the
University of Califor-nia, Los Angeles, in 1971.Since 1977 he has
been amember of the technical staffat Lincoln Laboratory, wherehe
has studied the detection,discrimination, and classifica-tion of
radar targets. He hascontributed chapters on sto-chastic observer
theory tovolumes 9 and 12 of the seriesAdvances in Control
Theory,and in 2000 he was elected aFellow of the IEEE.
. was an associate leader of theSurveillance Systems groupuntil
he retired from LincolnLaboratory in 1999. Hisbackground is in
radar-systemdesign and signal processing.His current interest is in
theidentification of movingtargets by using high-resolu-tion range
profiles and inverse-synthetic-aperture radar im-ages. He received
his B.S.E.E.degree from the University ofRhode Island and his
S.M.E.E.degree from MIT. He has beenat Lincoln Laboratory
since1971.