Air Force sbir 04. 1 Proposal Submission Instructions

1. Compact 1.55-micron fiber-optic coherent Doppler lidar for low-airspeed measurements, P. Mamidipudi, P. Gatchell, R. Changkakoti, P. Rogers, Optical Air Data Systems [5086-22], Laser Radar Technology and Applications VIII, 21–25 April 2003, Orlando, Florida, USA

KEYWORDS: Target Detection, Multi-Discriminate Sensing, Low-Cost Active/Passive EO Technology, Hardware and Signal Processing.

AF04-230 TITLE: Machine Learning for Robust Automatic Target Recognition
TECHNOLOGY AREAS: Sensors, Electronics, Battlespace
OBJECTIVE: The objective of this effort is to develop approaches to improve Automatic Target Recognition (ATR) system performance in operating conditions different than for which the system was trained. The resulting improved ATR system performance will improve time critical targeting and surveillance operations.
BACKGROUND: Simply stated, the purpose of an ATR system is to make correct decisions concerning target identification. The desired attributes of a good ATR are discrimination and robustness. That is, it should correctly classify a target of a particular type regardless of variations in the targets appearance or extended operating conditions (EOCs). EOCs might include: clutter and noise variations, minor design differences, pose, revetment, partial obscurations, and articulation of movable parts. At the same time, the ATR should reject out-of-class targets despite their similarities to in-class targets. Machine learning techniques could greatly improve performance in EOCs. For instance, perhaps an algorithm could be developed to automatically adapt thresholds based on learning of samples from several known operating conditions to scenarios with new operating conditions i.e. interpolate/extrapolate ATR parameters or select/adjust features based on estimates of current conditions. AFRL/SNA has been leading the research in model based ATR systems because the approach offers greater robustness to hard EOCs such as partial obscurations and part articulation. Model based ATR uses detailed phenomenon models of targets which can be articulated and varied as needed to predict measured data. One significant problem has been the mismatch between synthetic and measured data. Machine learning techniques might be able to develop mappings between synthetic and measure data so that synthetic driven model based ATRs could improve performance. Another issue is known as the feature aided tracking problem where the system seeks to assure that different looks at a target actually come from the same target. While identification may be impossible from single looks at a target, the system could still learn the target on the fly so that future looks are predictable and therefore can support tracker report association functions.
DESCRIPTION: The overall goal of this effort is to develop innovative machine learning algorithms and techniques that lead to improved ATR robustness. AFRL/SNA is especially looking for improvements to Model-Based ATR and feature aided tracking. Suggested problems to consider include but are not limited to:

1) Learning mapping between synthetic and measured radar data

2) Automated ATR parameter selection

3) Learning targets on the fly

A variety of sensor data may be considered; these include radar, video, IR, SAR, and laser. Preferably, initial efforts should focus on radar data such as 1D profiles, SAR, range-Doppler, and 3D imaging. Ideally, the techniques developed might expand to other sensors.
PHASE I: Develop an approach and prototype algorithms to improve ATR robustness. Develop arguments for feasibility of approach. This likely includes combinations of small demonstrations and theoretical arguments. Phase I proposals should identify data to be used and how progress will be measured.
PHASE II: Further develop the approach and algorithms identified in Phase I. Develop and demonstrate attributes not specifically demonstrated in Phase I. Conduct performance analysis and demonstrate utility to improve ATR robustness. Phase II proposals should identify data to be used and how progress will be measured.

DUAL USE APPLICATION: A wide of potential applications exist for automated target recognition including medical diagnosis, forensic analysis, and automated facial recognition applied to anti-terrorism screening.

RELATED REFERENCES;
1. http://www.mbvlab.wpafb.af.mil/paper.html has several dozen papers in the area. The first three references given below are available there and are specified here because they are most relevant.
2. J. C. Mossing, T. D. Ross, "An Evaluation of SAR ATR Algorithm Performance Sensitivity to MSTAR Extended Operating Conditions", pp.554-565, Proc. SPIE, vol 3370, Algorithms for Synthetic Aperture Radar Imagery V, April 1998
3. E. R. Keydel, S.W. Lee, J.T. Moore, "MSTAR extended operating conditions: a tutorial", Proc SPIE, Vol. 2757, pp.228-242, Algorithms for synthetic Aperture Radar Imagery IV, April 1997
4. T. D. Ross, L. Westerkamp, E. Zelnio, T. Burns, "Extensibility and other model-based ATR Evaluation Concepts" Proc. SPIE, Vol. 3070, pp.213-222, Algorithms for synthetic Aperture Radar Imagery IV, April 1997
5. Duda, R.O. and P.E. Hart, 1973. Pattern Classification and Scene Analysis, Wiley.
6. http://www.mbvlab.wpafb.af.mil/paper.html has several dozen papers in the area. The first three references given below are available there and are specified here because they are most relevant.
7. J. C. Mossing, T. D. Ross, "An Evaluation of SAR ATR Algorithm Performance Sensitivity to MSTAR Extended Operating Conditions", pp.554-565, Proc. SPIE, vol 3370, Algorithms for Synthetic Aperture Radar Imagery V, April 1998
8. E. R. Keydel, S.W. Lee, J.T. Moore, "MSTAR extended operating conditions: a tutorial", Proc SPIE, Vol. 2757, pp.228-242, Algorithms for synthetic Aperture Radar Imagery IV, April 1997
9. T. D. Ross, L. Westerkamp, E. Zelnio, T. Burns, "Extensibility and other model-based ATR Evaluation Concepts" Proc. SPIE, Vol. 3070, pp.213-222, Algorithms for synthetic Aperture Radar Imagery IV, April 1997
10. Duda, R.O. and P.E. Hart, 1973. Pattern Classification and Scene Analysis, Wiley.
11. http://www.mbvlab.wpafb.af.mil/paper.html has several dozen papers in the area. The first three references given below are available there and are specified here because they are most relevant.
12. J. C. Mossing, T. D. Ross, "An Evaluation of SAR ATR Algorithm Performance Sensitivity to MSTAR Extended Operating Conditions", pp.554-565, Proc. SPIE, vol 3370, Algorithms for Synthetic Aperture Radar Imagery V, April 1998
13. E. R. Keydel, S.W. Lee, J.T. Moore, "MSTAR extended operating conditions: a tutorial", Proc SPIE, Vol. 2757, pp.228-242, Algorithms for synthetic Aperture Radar Imagery IV, April 1997
14. T. D. Ross, L. Westerkamp, E. Zelnio, T. Burns, "Extensibility and other model-based ATR Evaluation Concepts" Proc. SPIE, Vol. 3070, pp.213-222, Algorithms for synthetic Aperture Radar Imagery IV, April 1997
15. Duda, R.O. and P.E. Hart, 1973. Pattern Classification and Scene Analysis, Wiley.
16. Goldberg, D. E., 1989. Genetic Algorithms in Search, Optimization, and Machine

Learning, Reading, MA: AddisonWesley.

IEEE SP.
18.Therrien, Charles W., 1989. Decision Estimation and Classification, Wiley.

KEYWORDS: Automatic Target Recognition, Model Based ATR, Machine Learning, High Range Resolution Radar, EO/IR, laser

AF04-231 TITLE: Modeling and Simulation Technologies for Multi-Sensor Dynamic Targeting
TECHNOLOGY AREAS: Sensors, Electronics, Battlespace
OBJECTIVE: Develop physics based modeling and simulation technologies for dynamic targeting applications.
DESCRIPTION: The Air Force Research Laboratory (AFRL) is investigating modeling and simulation technologies that could provide the Air Force with an improved capability to find, identify, and engage mobile targets. Modeling and simulation of widely varied scenarios and highly dynamic environments that accurately reflect what the sensor observes including any all derived target/object features is required to address this requirement. Dynamic scenario simulation and analysis techniques should support trade studies and analysis for Automatic Target Recognition (ATR) and sensor fusion algorithms and dynamic multi-sensor planning and exploitation.
The goal of this research is to develop simulations of dynamic target and background behavior and phenomenology to aid in the development of ATR and sensor fusion algorithms and the characterization of their performance, including the modeling of the ATR and sensor fusion algorithms/components/systems themselves. This topic solicits innovative solutions in the following areas: (1) Large scale scene modeling and simulation techniques. (2) Phenomenology prediction methods for dynamic targets in backgrounds (3) Modeling and simulation of algorithm performance characteristics (4) Methods for simulating ATR and sensor fusion performance in systems and family of systems context. Proposals that address one or a combination of these technology areas will be considered.
PHASE I: Address at least one of the following: (1) Develop a proof of concept for large scale dynamic scenario simulation that models platform/sensor interacting with the environment (2) Prototype physics-based modeling tool for a simple object or an object in terrain setting. (3) Develop a proof-of-concept software design document for ATR and sensor fusion performance estimation tools. (4) Develop a design for a simulation environment that enables interactive platform/sensor feedback and control analysis.
PHASE II: Develop and demonstrate at least one of the following: (1) Large Scale dynamic scene simulation that models platform/sensor interacting with the environment (2) Physics-based modeling tools for a complex moving object in a terrain setting (3) Software tools for efficient ATR and sensor fusion performance estimation. (4) Simulation environment that enables interactive platform/sensor feedback and control analysis.
DUAL USE COMMERCIALIZATION: A wide range of potential applications exist for large scene modeling such as oil and other natural resource exploitation, optimum use of agricultural assets, management of forestry resources, and demographic environmental impacts being prime examples.
REFERENCES:

1. Sullivan, D., D. Andersh, T. Courtney, N Buesing, and P. Jones, “Development of SAR Scene Modeling Tools for ATR Performance Evaluation,” Algorithms for Synthetic Aperture Radar Imagery VI , SPIE, Vol. 3271, April 1999, pp. 572 – 581.

KEYWORDS: Modeling and Simulation Techniques, Dynamic Targeting Applications, Detect and ID Mobile Targets, Sensor Fusion Algorithms.

AF04-233 TITLE: Multisensor Time Synchronization

AF04-234 TITLE: Algorithms for Real Time Corrections of Multi-Path Errors using Modernized GPS Signals

1. Multipath Error Reduction in Signal Processing, Alexey Zhdanov, Victor Veitsel, Mark Zhodzishsky and Javad Ashjaee, Javad Positioning Systems

2. Multipath Considerations for Ground Based Ranging Sources, C.G. Bartone, Ohio University

3. The Benefits of the GPS Three Frequencies on the Ambiguity Resolution Techniques, C. Bonillo-Martinez, M. Toledo-Lopez, M. Romay-Merino, GMV S.A.

4. Performance of GPS Receivers with More Than One Multipath, C. Macabiau, B. Roturier, A. Benhallam, CNS Research Laboratory of the ENAC, France; E. Chatre, STNA

5. GPS Carrier Phase Multipath Reduction Using SNR Measurements to Characterize an Effective Reflector, Reichert, P. Axelrad, University of Colorado

AF04-236 TITLE: Digital Array Analog to Digital Converter