Post on 06-Sep-2020
Lori WesterkampSensor ATR Technology Division
Sensors DirectorateAir Force Research Laboratory
Sensors Directorate and ATR Overviewfor Integrated Fusion, Performance Prediction, and Sensor
Management for ATE MURI
21 July 2006
2
AFRLMGen T. Bowlds
XPStaff
AirVehicles
SpaceVehicles Information Munitions Directed
Energy
AFOSRMaterials &
Manu-facturing
Sensors Propulsion HumanEffectiveness
AFRL Technology Directorates
3
Sensors Directorate
Our Mission
Robust sensors and adaptive countermeasures that guarantee complete freedom of air and space
operations for our forces, and deny these capabilities to our adversaries at times and places
of our choosing.
To lead the discovery, development, and integration of affordable sensor and countermeasure
technologies for our warfighters.
Our Vision
4
• Sense, identify, and track all air and surface targets and threats world wide and in all weather
• Counter “difficult” targets (WMD, hidden, LO)
• Protect air and space assets
• Control the battlespace electromagnetic spectrum
• Rapidly prosecute time-critical targets and threats
Priority Warfighter NeedsBeing Addressed by Sensors Directorate
Balance: Performance & Affordability
5
Sensors Directorate Technology Thrusts
Radio Frequency Sensors &
Countermeasures
Electro-Optical Sensors &
Countermeasures
Automatic Target Recognition
& Sensor Fusion
• Signatures & Modeling• Assessment &
Foundation• Innovative Algorithms
Enabling Sub-thrusts
• Radio Frequency Apertures• Algorithms & Phenomenology• Digital Receivers & Exciters• Reference Systems• Components
• Transmitters & Receivers
• Phenomenology &Algorithms
• Optical Apertures
Application Sub-thrusts • Battlespace Access• Persistent ISR of the Battlespace• Prosecution of Time Sensitive Targets
6
Potential Growth Areas
• Hidden Target Detection & Identification
• Automatic Target Recognition (ATR)& Sensor Fusion
• Electronic Warfare
• Time Critical Targeting (TCT)
7
ATR Scope
GEOLOCATE
FIND
TRACK
ID M60
FUN
CTI
ON
S
SENSORS
MATURATION PROCESSAlgorithms
Signatures
Assessment0 100
1
PD
FAR
8
Challenge ProblemsStandard MetricsATR Theory
Characterized PerformanceHigh Performance ComputingOperational Databases
ASSESSMENT& FOUNDATION
Sensor Data Management System (SDMS)
ATR Approach
INNOVATIVE ALGORITHMS
FIND FIX TRACK & ID
Phenomenology ExplorationEM ModelingSynthetic Data
Operational Target Models/Databases
SIGNATURES & MODELING
SignatureCenter
5
10
15
20
10
20
30
40
02.5
5
7.5
10
5
10
15
20
9
ATR Facilities
0 100
1
PD
FAR
PDPID
Sensor Resolution
Sensor Data Management System (SDMS)
Signature Center
Advanced Recognition Capability (ARC) Facility
10
Operating Conditions (OCs)
TargetsTargetsSensorsSensors
InteractionsInteractions
EnvironmentEnvironment
11
How can we make ATR easier?
• “Better” Features– Greater
discriminating power– Less variability
• “Better” Data– Resolution, dwell,
persistence, …– Multiple sources
(modes, looks, phenomenologies)
Often conflicting!
Fuel DrumsSide Fenders
Smoke Canisters
Length 6.9 mWidth 3.6 mHeight 2.2 m
Length 6.6 mWidth 3.5 mHeight 2.4 m
T-72 T-62
12
ATR-Driven Sensing
Cueing, Prioritization for the Human
~1’10’ ~1’
Short Circuits
Resonant Cross Slots Cavity
Low Dielectric/Low Loss Face Sheet
13
‘Clear Box’ View of ATR
0 100
1
PD
FAR
Environment
DetectDetectTrackTrack
GeolocateGeolocateIDID
SensorTargetATR
DecisionsHuman
Decisions
TrainedFeatures Templates Models
Target Knowledge
FeatureExtractor
Discriminator
DecisionRule
Target Knowledge
Target Models &Database
14
Multiple Sensor ATR
Target Models &DatabaseSensor
Model
Environment
DetectDetectTrackTrack
GeolocateGeolocateIDID
Sensor(s)TargetATR
DecisionsHuman
Decisions
Sensor Management Registration
Environment Model
Performance Model
Adaptation
Behavior Models
Anticipate
15
ATR Theory GoalsATR Theory Goals
• Performance Prediction– Algorithm (or feature) dependent – As a function of sensing, target, and environment conditions– Tool for developers (how will this new feature help?),
“maturers” (does this work well enough for my application?), and users (is this system going to work in my scenario today?)
• Performance Bounds– Algorithm independent performance limits– As a function of sensing, target, and environment conditions– Is the limit the algorithm (keep working) or the sensor
(motivate sensor improvements)?• Performance Models
– Packaged tools from Performance Prediction usable by sensor management and fusion algorithms to weigh sensor contribution
16
Growing ATR Complexity
Environments
Targets
Military Vehicles
Aircraft
Civilian Vehicles
Dismounts
Open Forested Urban
Function
Tougher ClutterMore Obscuration
SmallerMore Variable
Detect
Track
ID
More S
pecif
ic