Title: Automatic Target Recognition Using Passive Radar and a Coordinated Flight Model
1Automatic Target Recognition Using Passive Radar
and a Coordinated Flight Model
- Lisa M. Ehrman
- Advisor Aaron D. Lanterman
2What is ATR?
- Automatic Target Recognition (ATR)
- Ability to classify a target
- Two Schools of Thought
- Create images of the targets, which enable you to
conduct ATR - Conduct ATR directly on data, ie target RCS
- Our Approach
- Want to identify aircraft model
- Select Passive Radar as illuminator
- Use RCS to identify aircraft model
3What is Passive Radar?
- Active Radar
- System has transmitter and receiver
- Transmitter sends pulses which bounce off target
- Determine target position and velocity from
energy that bounces back
- Passive Radar
- System only has a receiver
- System exploits transmitters already available,
- ie TV, FM Radio used to determine target
position and velocity
4Why Use Passive Radar?
- Benefits
- Covert
- Cheap
- Not as susceptible to bad weather or stealth
- Lower frequencies ? RCS varies more slowly with
time - Challenges
- Transmitting signal not designed for target
detection tracking - Challenges have already been overcome by Howland,
Herman, Lockheed Martin,
5Radar Cross Section (RCS)
- RCS Depends on
- Incident Azimuth and Elevation
- Observed Azimuth and Elevation
AZI
AZO
ELI
ELO
PD
PR
Transmitter
Receiver
6Coordinated Flight Model
- Goal
- Estimate Aircraft Orientation from Position
- Key Parameters
- Heading
-
- Pitch
- Roll
- Assume Yaw 0
7Modeling RCS
- 1) Use FISC to Model RCS as function of
- Incident Azimuth and Elevation
- Observed Azimuth and Elevation
2) Use NEC2 to Model Receiving Antenna Gain
- 3) Use AREPS to Model Propagation Losses
- Between Transmitter and Aircraft
- Between Aircraft and Receiver
8Creating Noisy Profiles
- Problem
- In Absence of Real Data, Simulate the Power
Profile Arriving at the Receiver - Add White Gaussian Noise
- Assume phase is equally likely everywhere in
0,2p - Traces out a circle in the Complex Plane, whose
radius is the RCS magnitude - Model the thermal noise as normally distributed
white Gaussian noise, acting independently in the
Real and Imaginary directions - Add the noise to the profile using
9Computing Noise Power
- Noise Figure Vs. Noise Power
- Noise Figure Accounts for
- Out-of-Band Interference
- Direct-Path Interference
- Solution
- Sweep Noise Figure
10Identifying the Aircraft
- Use Loglikelihoods to Identify Aircraft Model
- Rician Model Compares the Library of Profiles to
the Simulated Profile at the Receiver - Treat each point in time as an independent sample
of a process - Loglikelihood is given by
- The aircraft with the largest loglikelihood is
matched to the target
11Geometries Tested
- Simple Scenarios
- Straight and Level Flight
- Banked Turn
- Complex Scenario
- Real Flight Profile
The flight pattern is a F-15C trajectory,
obtained from the Joint Helmet Cuing system,
mission JH-16, conducted by the 445th Flight
Test Squadron at Edwards Air Force Base in May
2000.
12Straight and Level FlightProbability of Error
Vs. Noise Figure
13Straight and Level FlightPower Profiles
14Straight and Level FlightConfusion Matrices
- Noise Figure 35 dB, Noise Power -166 dB
- Noise Figure 40 dB, Noise Power -161 dB
- Noise Figure 45 dB, Noise Power -156 dB
15Banked-Turn Flight ProfileProbability of Error
Vs. Noise Figure
16Banked-Turn Flight ProfilePower Profiles
17Banked-Turn Flight ProfileConfusion Matrices
- Noise Figure 45 dB, Noise Power -156 dB
- Noise Figure 50 dB, Noise Power -151 dB
- Noise Figure 60 dB, Noise Power -141 dB
18F-15 Trajectory 3-D
Z
Y
X
19F-15 Trajectory Top View
20F-15 Trajectory, Real AnglesProbability of
Error Vs. Noise Figure
21F-15 Trajectory, Real AnglesPower Profiles
22F-15 Trajectory, Real AnglesConfusion Matrices
- Noise Figure 55 dB, Noise Power -146 dB
- Noise Figure 60 dB, Noise Power -141 dB
- Noise Figure 65 dB, Noise Power -136 dB
23F-15 Trajectory, Est. AnglesProbability of
Error Vs. Noise Figure
24F-15 Trajectory, Est. AnglesPower Profiles
25F-15 Trajectory, Est. AnglesConfusion Matrices
- Noise Figure 55 dB, Noise Power -146 dB
- Noise Figure 60 dB, Noise Power -141 dB
- Noise Figure 65 dB, Noise Power -136 dB
26Estimating Aircraft Heading
Z
Y
X
27Estimating Aircraft Pitch
Z
Y
X
28Estimating Aircraft Roll
Z
Y
X
29Conclusions
- Performance Varies Greatly with SNR
- If you can get SNR 1, this is a viable approach
- If you have trouble correctly identifying
aircraft and SNR 1, you probably need a more
sophisticated means for estimating aircraft
orientation
30Future Work
- Finish the FISC Database, using more
sophisticated techniques - Determine whether or not the out-of-band and
direct-path interference can be accounted for in
this manner - Determine impact of errors in position estimates
31BACK-UP SLIDES
32SYSTEM DESCRIPTION
- Transmitter
- GPS Location 520100 N, 050300 E
- Altitude (ASL) 375 m
- Frequency 100 MHz
- Peak Power 100 kW
- Type Omni-directional
- Polarization Horizontal
- Receiver
- GPS Location 520636 N, 041926 E
- Altitude (ASL) 100 m
- Direction 320 (0N, 90E, 180S, 270W)
33RECEIVER GAIN PATTERN