ASU MAT 591 Opportunities in Industry ASU MAT

ASU MAT 591: Opportunities in Industry! ASU MAT 591: Opportunities In Industry! 1

ASU MAT 591: Opportunities in Industry! Advanced MTI Algorithms Howard Mendelson Principal Investigator 21 August 2000 2

ASU MAT 591: Opportunities in Industry! Problem Advanced MTI Algorithms l l SAR systems provide excellent intelligence concerning status of fixed installations (assuming no electronic countermeasures (ECM) are employed) Warfighter requires precise information describing MOVING formations of troops and weapons – Formations may be slow moving and thus difficult to distinguish from background clutter – Formations (as well as fixed targets) may be screened by ECM l l Our customers now specify high fidelity moving target indication (MTI) and fixed target indication (FTI) with interference rejection capabilities for their battlefield surveillance systems. These issues make it imperative for us to develop the techniques necessary to provide these capabilities 3

ASU MAT 591: Opportunities in Industry! STATE OF THE ART Advanced MTI Algorithms l l l DPCA – Not data adaptive ADSAR – Data adaptive but not jammer resistant SPACE TIME ADAPTIVE PROCESSING (STAP) – No Fielded GMTI Systems – Computationally Intensive – Traditional SMI Approach Produces Large Numbers of False Alarms 4

ASU MAT 591: Opportunities in Industry! Approach Advanced MTI Algorithms l Develop Post Doppler Eigenspace Analysis Techniques – Advantages § § § Lower false alarm rate than traditional SMI approach Simultaneous SAR and MTI in the presence of ECM Common processing framework for clutter and jammer suppression Higher Signal-to-Background Ratio (SBR) after interference suppression Smaller training data set required for STAP algorithms Computational Efficiency 5

ASU MAT 591: Opportunities in Industry! Advanced MTI Algorithms Sample Matrix Inversion (SMI) Interference Suppression Algorithm Input Data (N channels) Form Covariance Estimates Invert Covariance Matrix Beamform Apply Inverse Detection Processing 6

ASU MAT 591: Opportunities in Industry! Advanced MTI Algorithms Eigendecomposition Interference Suppression Algorithm Input Data (N channels) Form Covariance Estimates Perform Eigendecomposition Determine No. of Interference Sources Project Data Orthogonally to Interference Subspace Beamform Detection Processing 7

ASU MAT 591: Opportunities in Industry! Covariance Estimation Channel 1 N/2 Rng Cells Guard Cells Cell of Interest Guard Cells N/2 Rng Cells Channel 2 N/2 Rng Cells Guard Cells Cell of Interest Guard Cells N/2 Rng Cells X 1 Channel N N/2 Rng Cells Guard Cells Cell of Interest Guard Cells N/2 Rng Cells No. of range cells used for Eigen processing is typically 1. 5 x No. of channels (Higher for SMI) Covariance estimate is computed in sliding window at every pixel No. of guard cells depends on range resolution . . . XN H is complex conjugate transpose 8

ASU MAT 591: Opportunities in Industry! Weight Calculation (SMI) Sample Matrix Inversion (SMI) subject to R$ C f w Sample Covariance Matrix Constraint Matrix Coefficient Vector Weight Vector H Hermitian adjoint (conjugate transpose) 9

ASU MAT 591: Opportunities in Industry! Weight Calculation (MNE) Minimum Norm Eigencancler (MNE) subject to Q r and Matrix of eigenvectors of estimated covariance matrix associated with interference C Constraint Matrix f Coefficient Vector w Weight Vector 10

ASU MAT 591: Opportunities in Industry! LM M&DS – ISRS IR&D SAR Testbed 24” 7” adjustable Channel 0 Receive Channel 1 Transmit/Receive Channel 2 Receive flight 11

ASU MAT 591: Opportunities in Industry! Controlled Mover in Clutter (Eigendecomposition) Advanced MTI Algorithms Controlled Moving Target 12

ASU MAT 591: Opportunities in Industry! Controlled Mover in Clutter (SMI) Advanced MTI Algorithms 13

ASU MAT 591: Opportunities in Industry! PRI Stagger Algorithm Advanced MTI Algorithms FFT Elements (or beams) 1 2 3. . . P-1 P FFT S T A P FFT 1 2 3. . . P-1 P FFT 14

ASU MAT 591: Opportunities in Industry! Covariance Estimation Channel 1 Stagger 0 N/2 Rng Cells Guard Cells Cell of Interest Guard Cells N/2 Rng Cells Channel 2 Stagger 0 N/2 Rng Cells Guard Cells Cell of Interest Guard Cells N/2 Rng Cells X 10 n Channel L Stagger Nstg - 1 N/2 Rng Cells Guard Cells Cell of Interest Guard Cells N/2 Rng Cells No. of range cells used for Eigen processing is typically 1. 5 x No. of channels x No. of staggers (Higher for SMI) Covariance estimate is computed in sliding window at every pixel No. of guard cells depends on range resolution . . . XLNstg-1 n H is complex conjugate transpose 15

ASU MAT 591: Opportunities in Industry! Data Collect Radar Image Tactical Targets 16

ASU MAT 591: Opportunities in Industry! Data Collect Tactical Targets Unprocessed Image SMI Processing Eigendecomposition Processing 17

ASU MAT 591: Opportunities in Industry! CFAR DETECTORS (GMTI) Adaptive Matched Filter (SMI) H 1 > < a AMF H 2 Generalized Likelihood Ratio Test (SMI) H 1 > < a GLRT H 2 Eigendecompsition Likelihood Ratio Test H 1 > < a PC H 2 18

ASU MAT 591: Opportunities in Industry! Detection Performance (Pfa = 10 -6 ) Unprocessed Image SMI - GLRT Detection Reports SMI - AMF Detection Reports LRT - Eigendecomposition Detection Reports 19

ASU MAT 591: Opportunities in Industry! Detection Performance Pfa = 10 -6 Unprocessed Image SMI - GLRT Detection Reports SMI - AMF Detection Reports LRT - Eigendecomposition Detection Reports 20

ASU MAT 591: Opportunities in Industry! RELOCATION ALGORITHM Uses Channel-to-Channel Phase Differences to Obtain Target Direction of Arrival (DOA) l Originally Developed for Three Channel “Uniformly” Spaced Array Without PRI Stagger l Assumed Clutter as only Interference Source l – Insufficient number of degrees of freedom available to deal with more than one interfering source l Can be extended – No. of channels greater than 3 – Multiple interfering sources – Non-uniform spacing 21

ASU MAT 591: Opportunities in Industry! RELOCATION ALGORITHM Assumed Signal Model 22

ASU MAT 591: Opportunities in Industry! RELOCATION ALGORITHM y 1 = y 2 = r r ( e$ , s ) @ ( e$ , s ) r r ( e$, s ) @ ( e$, s ) 1 2 1 Tgt 2 Tgt e$= First eigenvector orthoganal to clutter direction 1 e$= Second eigenvector orthoganal to clutter direction 2 Same eigenvectors computed during interference suppression and detection processing Phase of target vector can now be found by solving for roots of quadratic Solution which provides largest return after beamforming is assumed correct 23

ASU MAT 591: Opportunities in Industry! Relocation Algorithm - Example Relocated Targets Original Target Detections 24

ASU MAT 591: Opportunities in Industry! RELOCATION ALGORITHM - 2 Assumed Signal Model Complex images from each channel are assumed to have been relocated to a common point 25

RELOCATION ALGORITHM - 2 (cont. ) ASU MAT 591: Opportunities in Industry! y 1 = y 2 = r r ( e$ , s ) @ ( e$ , s ) r r ( e$, s ) @ ( e$, s ) 1 2 1 Tgt 2 Tgt e$= First eigenvector orthoganal to clutter direction 1 e$= Second eigenvector orthoganal to clutter direction 2 Same eigenvectors computed during interference suppression and detection processing Phase of target vector can now be found by solving for roots of quadratic Solution which provides largest return after beamforming is assumed correct 26

ASU MAT 591: Opportunities in Industry! Geolocation Accuracy Cramer Rao bound of interferometer measurement accuracy used to estimate cross range error 27

ASU MAT 591: Opportunities in Industry! Target Reports Known Targets SMI based STAP Eigenanalysis based STAP 28

ASU MAT 591: Opportunities in Industry! Target Reports Original Detections Relocated Targets Unprocessed Target Detections Relocated Target Detections 29

ASU MAT 591: Opportunities in Industry! Multi-Stage False Alarm Reduction Processing Multichannel Complex Image Data Covariance Estimate Find Eigenvalues and Eigenvectors Find Noise Subspace Dimension Form Image Projections Produce Low Resolution SAR Image Form Interference Suppression Projections Produce Interference Suppressed Data Field Form Estimated Steering Vector Compute AOA (Radial Speed) Estimates Perform CFAR Thresholding Compute Cancellation Ratios of Threshold Crossings Determine AOA Consistency of Estimates of Possible Detections Detection reports Location, Speed and Heading Estimates 30

ASU MAT 591: Opportunities in Industry! SUMMARY l Multiple post-Doppler STAP algorithms studied and evaluated for clutter suppression and target detection – Eigenanalysis, SMI – Single Doppler bin, adjacent Doppler bin, PRI stagger l l “Mono-pulse” location algorithm developed and tested on collected data Work ongoing to develop algorithm upgrades 31