Sequential Adaptive MultiModality Target Detection and Classification using

Sequential Adaptive Multi-Modality Target Detection and Classification using Physics-Based Models • Professor Andrew E. Yagle (PI) (EECS) Mine detection, channel identification • Professor Alfred O. Hero III (EECS) Sensor scheduling, nonparametric statistical models • Professor Kamal Sarabandi (Director, Rad Lab) Vehicle and foliage radar physics-based modeling

Redu ced M SM Da ta ted ula ta Performance Da Sim al Sc m ena od r els ios /S odels Sc m ena od ri els os /S en so r tu TUT Ac en s or o M d uce d e R dels Matrix Hybrid UXO simulated/real data Environment Modification

Research Project Objectives 1. Develop physics-based and statistical models for radar scattering from vehicles under foliage; 2. Develop physics-based and statistical models for GPR and EMI responses of buried land mines; 3. Develop statistical algorithms for detection and sensor management for multiple sensor modalities; 4. Insert physics-based models into statistical sensor management and detection multi-modal algorithms; 5. Evaluate the resulting procedure on realistic models (physics-based and statistical) and on real data.

1. Physics-based models for vehicles under foliage • Sarabandi: Developed very realistic models for radar scattering from vehicles under foliage. Significance: Use in Monte Carlo simulations to develop statistical models to be inserted into sensor management and detection algorithms. Significance: Evaluation of resulting algorithms. • Hero: Non-parametric model for scattered field. Significance: Monte Carlo takes forever. So developed statistical interpolation algorithm.

2. Physics-Based Models for Land Mines Yagle: Applied time-reversal imaging to GPR. Significance: Basic mine detection algorithm. Yagle: Developed metal detector model for mine detection and decay rate to distinguish Fe and Al. Significance: Quick way to develop multiple modalities and evaluate them on real test data. Significance: Insert these into Hero’s statistical sensor scheduling and detection algorithms.

3. Statistical Sensor Management and Detection • Hero: Myopic and non-myopic multi-sensor management for detection and tracking. Significance: Greatly reduces computation. Demonstrated tracking of dozens of actual targets. • Hero: Aggregation strategy for objective function maximization for distributed sensors. Significance: Easier to find global maximum for optimal sensor scheduling and deployment.

4. Inserting Physics-Based Statistical Models into Statistical Sensor Management Algorithms • Sarabandi, Hero, Yagle: In Years #3 and #4 we will insert statistical models developed using Monte Carlo simulations on realistic physics-based models into statistical sensor management and detection algorithms. Significance: This is the heart of our MURI, where payoff comes: Rigorous statistical algorithms for sensor management based on realistic physics-based and statistical models.

5. Evaluation of Algorithms We Have Developed • Sarabandi: Using realistic physics-based models we have now developed for vehicles under foliage, we can now evaluate algorithms in a wide variety of physical scenarios in which ground truth can be set. • Real data also to be used, but real and multi-modal data only available for limited set of scenarios. • Hero: Statistical algorithms used to evaluate the performance of sensor management algorithms. • Yagle: Similarly for mine detection problem.

New. Research Project Objectives • Develop physics-based and statistical models for radar scattering from vehicles under foliage; • Develop physics-based and statistical models for GPR and EMI responses of buried land mines; • Develop statistical algorithms for detection and sensor management for multiple sensor modalities; • Insert physics-based models into statistical sensor management and detection multi-modal algorithms; • Evaluate the resulting procedure on realistic models (physics-based and statistical) and on real data.

Old Research Project Objectives • Develop overall algorithm for sequential detection, sensor management & selection • Develop physics-based models for mines and vehicles; statistical models from them. • Simplify physics-based models using functional-analysis-based approximation • Evaluate the resulting procedure on realistic models (statistical simulations) and real data

Changes to Original Program Objectives: • • • PROJECT OBJECTIVES DROPPED: Optimal basis function representations Basis-function-based inverse scattering Underground structures PROJECT OBJECTIVES ADDED: More specific about physics-based models leading to statistical models, to be inserted in statistical sensor management algorithms

ALGORITHM DEVELOPMENT SENSOR MANAGEMENT VEHICLE DETECTION FOLIAGE MINE DETECTION TANK GPR FREQ. POLAR FREQ. . METAL DETECTOR

Overall Algorithm: Overview Target detector/ classifier

GOAL #1: ACCOMPLISHMENTS I • Performed phenomenological studies of: (a) physics-based clutter models (b) physics-based target models Significance: Basic understanding of effects is vital for interpreting results. These proved very useful in developing the statistical models of scattered fields.

GOAL #1: ACCOMPLISHMENTS II • New results: Target-clutter interaction; Multiple scattering from needle clusters; Closed-form solution for scattering from a disk of arbitrary shape. • Developed time-reversal method for foliage camouflaged target detection. Significance: This is one of the physicsbased models for detection (project title).

GOAL #1: ACCOMPLISHMENTS III • Developing iterative frequency-correlation based forest radar channel identification; New approach for attenuation estimation. Significance: Procedure for deconvolving effects of propagation through foliage. • Developing iterative physical optics approach to account for foliage shadowing. Significance: Greatly reduces computation

GOAL #2: ACCOMPLISHMENTS I • Developed mine detection algorithm from SAR GPR using range migration imaging (with Jay Marble). Significance: Physics-based algorithm for imaging mines from ground-penetrating radar (project title). • Developed 2 D and 3 D blind deconvolution algorithms for radar channel identification (with Siddharth Shah). Significance: Apply to blind deconvolution of channel propagation effects for mines, and perhaps for foliage.

GOAL #2: ACCOMPLISHMENTS II • Developed “hyperbola-flattening transform” algorithm for feature detection in GPR data. Significance: Preliminary detection stage using less computation than range-migration imaging • Working on material discrimination using decay rates from magnetometer (metal detector) data. Significance: Multi-modal mine detection.

GOAL #3: ACCOMPLISHMENTS I • Developed non-parametric statistical modelling of scattered fields using Markov random fields. Significance: Can model scattered fields from a few observations, extrapolating the rest. This saves much time in Monte Carlo statistical model development. • Developed target model reduction technique. Significance: Can model vehicles using a lower dimensional manifold, simplifying detection.

GOAL #3: ACCOMPLISHMENTS II • Developed myopic distributed multi-sensor multilook detection and tracking sensor management algorithms using Renyi divergence. Significance: Optimal sensor scheduling too hard when many targets and sensors present. Particle filtering + Renyi-divergence-based scheduling reduces complexity. Tracking of dozens of actual targets was demonstrated.

GOAL #3: ACCOMPLISHMENTS III • Developing non-myopic distributed multi-sensor multi-look detection and tracking sensor management algorithms using queue learning. Significance: Can develop quantification of the performance-vs-complexity tradeoff issue. • Developing aggregation strategies for distributed sensors and quantified performance tradeoffs. Significance: Allows fast computation of maxima of objective functions; these dictate strategies.

Synergistic Activities: Hero General Dynamics (formerly Veridian, ERIM): C. Kreucher: sensor management & scheduling K. Kastella: sensor management J. Ackenhusen: mine detection ARL: NAS-SED review panel member N. Patwari (student) summer internship G. Shih (student) summer internship

Synergistic Activities: Sarabandi General Dynamics: John Ackenhusen BAE: Norm Byer FCS COMMUNICATIONS: Jim Freibersiser (DARPA PM) Barry Perlman (CECOM) ARL: Ed Burke (mm wave), Brian Sadler, Bruce Wallace

Synergistic Activities: Yagle General Dynamics (formerly Veridian, ERIM): Jay Marble, student (ARO mine research) Brian Fischer, student (Low RCS material design) Chris Wackerman, former Ph. D. student

SEQUENTIAL ADAPTIVE MULTI-MODALITY TARGET DETECTION AND CLASSIFICATION USING PHYSICS-BASED MODELS (PICTURE) APPROACH • Develop realistic physics-based models; • Perform statistical simulations to obtain distributions of measured scattered fields; • Develop sensor scheduling and detection algorithms using these statistical models; • Evaluate algorithms using statistical measures; • Apply algorithms to real multi-modal data. ARMY COLLABORATIONS • Army Night Vision Lab (GPR & IR mine field data). OBJECTIVES • Develop algorithms for detection of landmines and tanks under trees using radar and IR sensors; • Develop data-adaptive algorithms for sensor scheduling and multi-modal sequential detection; • Evaluate the algorithms using Monte Carlo type simulations on realistic models, and on real data. ARMY RELEVANCE • Detection of landmines and tanks under trees has obvious Army relevance ACCOMPLISHMENTS • Phenomenological studies of radar clutter and target+clutter using realistic physics-based models; • Developed non-parametric MRF models for these; • Developed myopic sequential adaptive sensor management algorithm for tracking problems; • Developed migration (time-reversal) algorithm for imaging land mines and evaluated on real GPR data. TRANSITION TO ARMY/ INDUSTRY • None yet.