Detection of Obscured Targets: Signal Processing James McClellan and Waymond R. Scott, Jr. School of Electrical and Computer Engineering Georgia Institute of Technology Atlanta, GA 30332-0250 jim.mcclellan@ece.gatech.edu 404-894-8325 Outline Introduction Location with Acoustic/Seismic Arrays Maneuvering Array (3x10) Cumulative Array strategy for imaging GPR and Quadtree Region Elimination Accomplishments/Plans MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 2
Three Sensor Experiment Sensor Adjustments and Features Adjustable Parameters for all three sensors Frequency range Frequency Resolution Spatial Resolution Integration time/bandwidth Height above ground Location Possible Features for sensors EMI Relaxation frequency Relaxation strength Relaxation shape Spatial response GPR Primary Reflections Multiple Reflections Depth Spatial Response Seismic Resonance Reflections Dispersion Spatial response MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 3 Multi-Resolution Processing GPR Imaging Quadtree Imaging @ increasing Resolution EMI SigProc (Eliminate Areas) Features Features Decision Process Exploit Correlation ID Detect Classify Seismic Ad-Hoc Array Imaging Features Training Target Localization @ specific sites MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 4
Outline Introduction Location with Acoustic/Seismic Arrays Maneuvering Array (3x10) Cumulative Array strategy for imaging GPR and Quadtree Region Elimination Accomplishments/Plans MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 5 Seismic Sensor Signal Generator Radar: R.F. Source, Demodulator, and Signal Processsing S N S Waveguide E.M. Waves Elastic Wave Transducer Elastic Surface Wave Mine Displacements Air Soil MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 6
Ground Contacting Seismic Sensor Array Deployed on a Small Robotic Platform Source Platform Sensor Platforms MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 7 Home in on a Target Problem Statement: Use a small number of source & receiver positions to locate targets, i.e., landmines Minimize the number of measurements Three phases 1. Probe phase: use a small 2-D array (rectangle or cross) Find general target area by imaging with reflected waves 2. Adaptive placement of additional sensors Maneuver receiver(s) to increase resolution Use Theory of Optimal Experiments 3. On-top of the target End-game: Verify the resonance MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 8
Adaptive Sensor Placement Additional sensors are added to the cumulative array 3 On-top for resonance 2 Target 1 Source Probe Array Probe Array finds general target area MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 9 Review from January Probe Phase Pick five sensors in a cross pattern Apply Imaging algorithm and plot the surface over a search grid Maneuver Phase Add one more sensor depending upon which direction to move Increase aperture of triangulation Spatial resolution increases which narrows down the area in which to search for target MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 10
TS-50 (1cm) Source at (-20,50) MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 11 Next Measurement 1 2 3 MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 12
Choose Next Sensor Position Want to formulate an optimal maneuvering strategy Instead of adding One Sensor, move the Whole Array to the next optimal position Append the array data at the new position with previous data Estimate the target location Find the best next sensor position Repeat this until target is localized MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 13 Steps in Optimal Maneuver MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 14
Steps in Optimal Maneuver Wave Separation by Prony Imaging algorithm - Data Model - ML solution for target position estimates - Performance bounds for position estimates Next optimal Array Position - D-optimal Design - Constrained Optimization MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 15 Data Model P-element array with position of every sensor is given in terms of array center Seismic wave penetration depth depends on frequency, hence a band of frequencies is used in processing K near-field wideband targets In the DFT, select L Frequency components Goal: estimate location (p) of the targets from measured array data MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 16
Data Model MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 17 Target Position Estimate The Cramer-Rao lower bound (CRLB) provides a lower bound for the variances of the unbiased estimators CRLB requires the inverse of the Fisher information matrix MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 18
Fisher Information Matrix for Position Estimate MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 19 Theory of Optimal Experiments The theory of optimal experiments uses various measures of the Fisher information matrix to produce decision rules Determinant Trace Maximum value along the diagonal D-optimal Design uses the determinant X. Liao and L. Carin, Application of the Theory of Optimal Experiments to Adaptive Electromagnetic-Induction Sensing of Buried Targets,'' IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 26, no.8, pp. 961 972, August 2004. MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 20
Next Optimal Position MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 21 Unique Problems for Seismic Target position is estimated from reflected seismic energy Source is nearby Need separation between forward and reverse waves Array position has to be between source and targets always Landmines reflections are not omni-directional Next array position has to be Constrained Between source and estimated target position Two ways to implement constrained optimization Circle constraint, or Penalty Function MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 22
Constrained optimization for next array position MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 23 Experimental Data Setup Single TS-50 landmine is buried at 1cm Reflected waves are separated at each position by using a line array of 15 sensors (3cm apart) Actual receiver is a single sensor (Synthetic Array) Measurements are grouped into line arrays From each line array, three sensors at equal distance positions are chosen With three line arrays, an array of nine sensors is available for 2-D imaging Inverse of the cost function is plotted for imaging algorithm MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 24
Seismic Detection of VS-1.6 AT Landmine Buried 5cm Deep in Experimental Model Landmine (22.2 cm diameter) buried 110 cm from first measurement location in 171 cm linear scan. 5 measurements with center line of sensor array along a line over the burial location. Compressional, Rayleigh Surface, and Reflected Waves. Resonance of Landmine-Soil System. Interactions of incident waves with buried landmine evident in measured data. MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 25 Wave Separation at each iteration: along top-most line array 1 2 3 4 MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 26
Data Movies All wave types Only the reflected wave After Prony processing MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 27 First Iteration ACTUAL TARGET : (135,135) ± 5 ESTIMATE (102,120) MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 28
Next Array Position Circle constraint R=30cm Movement Penalty MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 29 Next array position values calculated on half circle of radius=30cm Values calculated on half circle MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 30
Second Iteration ACTUAL TARGET : (135,135) ± 5 ESTIMATE: ALONE (100,127), CUM(112,127) ALONE CUMULATIVE MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 31 Next Array Position Values calculated on half circle MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 32
Third Iteration ACTUAL TARGET : (135,135) ± 5 ESTIMATE: ALONE (143,138), CUM(127,133) ALONE CUMULATIVE MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 33 Fourth Iteration ACTUAL TARGET : (135,135) ± 5 ESTIMATE: ALONE (124,144), CUM(146,139) ALONE CUMULATIVE MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 34
All Four Iterations: Imaging with one array at a time MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 35 Four Iterations: Cumulative Imaging MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 36
2D array (3 X 10): Implementation Three lines having 10 sensors each Sensors are ground contacting accelerometers LabView is used to control the movement of array and seismic source firing and interface with MATLAB Processing algorithms are implemented in MATLAB Target is a VS1.6 mine buried at 5cm MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 37 Experimental Setup of Seismic Landmine Detection Measurements 3 by 10 Element Array of Ground- Contacting Sensors Seismic Source 174 cm 45 cm VS-1.6 AT Landmine (5cm deep) 50 Tons of Damp, Compacted Sand Scan Region of 2m by 2m MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 38
Demo in Sandbox Target is localized in four iterations Each iteration takes 10 secs. for processing and 30 secs. for data acquisition Total time for four iterations is 2 minutes Typical raster scan takes a few hours to locate the target with a large number of measurements EXPERIMENTAL SETUP MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 39 Four Iterations MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 40
Four Iterations: another run MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 41 End-game Detection We can further scan in line with the last estimated target position Extract the reverse wave and try to locate the exact location of resonance Movies shown the extracted wave with the line array of 30 sensors, which is moved toward the target with 1cm increment Target center is (50,50) and the starting location and position of target is shown MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 42
Movie: VS-1.6 (5cm) Extracted reverse wave as array moves near and above the target MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 43 Movie: TS-50 (1cm) Extracted reverse wave as array moves near and above the target MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 44
Outline Introduction Location with Acoustic/Seismic Arrays Maneuvering Array (3x10) Cumulative Array strategy for imaging GPR and Quadtree Region Elimination Accomplishments/Plans MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 45 Detecting non-target areas by GPR Θ : Beamwidth of the transmitter and receiver antennas d tr : Transmitter Receiver Distance h : Antenna Height If there is no detection at this location Shaded region does not contain targets Move transmitter by Instead of one step size. 2h tanθ d tr MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 46
Results from the GPR Air Measurements Measured Data Theoretical Reference Signal Shotput in Air; Measured Data Double Differentiated Gaussian Pulse 100 200 300 400 500 600 700 800-80 -100-120 -140-160 -180 0.15 0.1 0.05 0-0.05 900 Threshold : 10 20 30 40 50 60 70 80 90 γ -200 2 1 = σ ε Q ( P FA P FA is a user defined parameter σ 2 is estimated from the data ) -1-0.5 0 0.5 1 time(seconds) x 10-9 Correlate with a delayed and scaled version of the transmitted signal Estimate the variance by taking average energy of the signals from non-target areas MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 47 Detection Test for first measurement point Detection Test 12 x 10-8 10 Signal Correlation Threshold 1.5 x 10-7 Detection Test at X = -54cm Signal Correlation Threshold 8 1 6 4 0.5 2 0 0-2 0 200 400 600 800 1000 1200 1400 1600 1800 2000 No detection, move 2h tan θ d tr -0.5 12 x 10-8 Detection Test at X = -72cm 10 Signal Correlation Threshold -1 0 200 400 600 800 1000 1200 1400 1600 1800 2000 8 6 4 2 0-2 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Target Detected! Apply time delay difference to find the apex of the hyperbola. MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 48
Monostatic Antenna in Homogeneous Medium t α+1 Δ t α z 0 2 2 2 2 4 2 2 tα = z0 + ( αδ) tα = ( z 2 0 + ( αδ) ) c c 2 2 2 2 4 2 2 2 tα + 1 = z0 + (( α + 1) Δ) tα + 1 = ( z 2 0 + ( αδ) + (2α + 1) Δ ) c c 2 2 4 2 tα + 1 tα = (2α + 1) Δ 2 Does not depend on target depth! c Find α which indicates how many step size the antenna is away from the target position MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 49 Target Position Found 4 x 10-5 Reading at X = -54 3 2 1 0-1 t α -2-3 0 100 200 300 400 500 600 700 800 900 1000 5 x 10-5 Reading at X = -52 4 TDD α = 26 is found and the antenna is moved to target position which is X = 0 3 2 t α+1 1 0-1 -2-3 0 100 200 300 400 500 600 700 800 900 1000 MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 50
Accomplishments Developed three sensor experiment to study multimodal processing Developed new metal detector and a radar Investigated three burial scenarios Showed responses for all the sensors over a variety of targets Demonstrated possible feature for multimodal/cooperative processing Developed new 3D quadtree strategy for GPR data Developed seismic experiments, models, and processing Developed optimal maneuver algorithm to locate targets with a seismic array Demonstrated reverse-time focusing and corresponding enhancement of mine signature Demonstrated imaging on numerical and experimental data from a clean and a cluttered environment Modified time-reverse imaging algorithms to include near field DOA and range estimates. The algorithms are verified for both numerical and experimental data with and without clutter. Modified wideband RELAX and CLEAN algorithms for the case of passive buried targets. The algorithms are verified for both numerical and experimental data with and without clutter. Developed a vector signal modeling algorithm based on IQML (Iterative Quadratic maximum Likelihood) to estimate the two-dimensional ω-k spectrum for multi-channel seismic data. Developed multi-static radar Demonstrated radar operation with and without clutter objects for four scenarios Investigated pre-stack migration imaging of multi-static data Buried structures Developed numerical model for a buried structure Demonstrated two possible configurations for a sensor Made measurement using multi-static radar MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 51 Plans Three sensor experiment (Landmine) Incorporate reverse-time focusing and imaging Incorporate multi-static radar More burial scenarios based on inputs from the signal processors Look for more connections between the sensor responses that can be exploited for multimodal/cooperative imaging/inversion/detection algorithms Imaging/inversion/detection algorithms Use reverse-time ideas to characterize the inhomogeneity of the ground Investigate the time reverse imaging algorithm for multi-static GPR data. Investigate the CLEAN and RELAX algorithms for target imaging from reflected data in the presence of forward waves with limited number of receivers. Develop elimination and end-game strategies for seismic detection with a maneuvering array Investigate joint imaging algorithms for GPR and seismic data. Buried Structures & Tunnels Experiments with multi-static radar Develop joint seismic/radar experiment Signal Processing MURI Review 3-Aug-05 Scott/McClellan, Georgia Tech 52