1. Lead ~ Discover ~ Develop ~ Deliver Air Force Research Laboratory Multimodal Data and Anomaly Detection in SSA at AMOS 15 Oct 2012 Dr. Keith Knox Air Force Maui Optical & Supercomputing Site Maui, Hawaii Photo of Briefer

2. Air Force Maui Optical & Super Computing Site 2

3. Air Force Maui Optical & Supercomputing Site Largest telescope in Department of Defense with 3.6m primary telescope Highest resolution adaptive optics in Department of Defense Largest electro-optical tracking facility in the Pacific 1960:The Advanced Research Projects Agency (ARPA) Midcourse Optical Station 1963:Site construction started by ARPA 1994:High Performance Computer Center (HPCC) completed 1999:Advanced Electro-Optical System (AEOS) completed 2001:Air Force Research Laboratory 50 Years of Service to the Department of Defense 3

4. Air Force Research Laboratory Directed Energy Directorate Maui Space Surveillance System MSSS (AFRL) 3.6m and 1.6m telescopes Maui Space Surveillance System High-resolution Imaging Orbital Tracking Space Object Characterization GEODSS 1.6m 4 AEOS 3.6m AEOS 3.6m GEODSS ( AF Space Command ) Ground-based Electro-Optical Deep Space Surveillance

5. 3.6-meter Telescope Advanced Electro-Optical System AEOS 3.6 Telescope 5

6. 1.6-meter Telescope 1.6 Meter Telescope Inside Dome 6

7. Adaptive Optics Imaging AEOS Visible Imager Hubble Space Telescope Adaptive Optics (AO) plus multi-frame blind deconvolution processing Terminator Imagery Day Night Terminator 7

8. Long-Wave Infrared Imaging Resolved thermal images Virtually diffraction-limited Nighttime Imagery Day Night Terminator AEOS Infrared Imager 8

9. Speckle Imaging Daytime Imagery Terminator Imagery Speckle Imaging 1.6m Raw data Processed Result Day Night Terminator 9

10. As satellite image size decreases … –Smaller satellite –Greater distance the satellite becomes completely unresolved –Satellites in geo orbit –Cubesat-class satellites Non-imaging Characterization 10

11. Non-Imaging Technique Temporal filter photometry –Measured brightness as function of time For an object facet to contribute to signal –Facet must be illuminated by Sun –Facet must be visible to sensor Sensor requirements are simple –Calibrated light bucket Temporal photometry 11

12. Astrodynamics & Tracking High Performance Computing Software Applications Institute for Space Situational Awareness The Institute meets these challenges by bringing together: Supercomputing expertise World-class researchers from AFRL Good orbit knowledge and some status info Accurate orbits and uncertainty Object identification status and health Accurate orbits and uncertainty Object identification status and health Now 15k Objects Future 150k Objects 12

13. Anomaly Detection and Multimodal Data in Astrodynamics 13

14. Space Catalog Anomalies 22,000 objects in the catalog What is an anomaly? –New satellite is launched –Debris is created –Satellite maneuvers –Object drifts –Satellite status has changed Track 22K objects –Look for deviations Paul Schumacher –“The Future Space Catalog” 14

15. Multimodal Space Catalog: Radar vs. Optical RA 1 Dec 1 RA 2 Dec 2 RA 3 Dec 3 3 observations ( x y z v x v y v z ) 6 scalars Range 2 Az 2 Elev 2 Range 1 Az 1 Elev 1 2 observations ( x y z v x v y v z ) 6 scalars 2 radar observations 3 optical observations Radar and Optical Observations translated into 3-D spatial coordinates 15

16. Observation Association is Hard For one object, all observations are connected SST 16 ? ? ? ? Space Surveillance Telescope For many objects, all observations are connected

17. Multimodal Data in Speckle Imaging 17

18. Speckle Imaging using Short Exposure Sequences High resolution details are lost in long exposures through the atmosphere: However, detail is encoded in short exposure images: Assume that target is constant over period of a few seconds. Then image reconstruction is possible: 18

19. Speckle Imaging Multi-Frame Blind Deconvolution (MFBD) o(x) True object Restored object MFBD Processing Minimize this cost function with respect to,, …, : … Noisy and blurred images … Blurring functions 19

20. Multimodal Data Improves Image Reconstruction Each wavelength experiences ~same optical path difference (OPD) due to atmospheric turbulence Wavefront phase is θ λ = OPD × 2π/λ Longer wavelength Shorter wavelength OPD in telescope pupil Infrared images define OPD, which in turn improves visible reconstruction Brandoch Calef, “Wavelength Diversity” 20

21. Multimodal Data in Non-Imaging 21

22. Spectrophotometry with BASS at AEOS IR spectrophotometry in 3-13.5 mm range Princeton CCD camera & filter collect images simultaneously with IR spectra 22

23. Modeling Reflected & Emitted Radiation Modeling space debris to match simultaneous IR and visible response HAMR objects –High Area-to-Mass Ratio Mark Skinner –“Fusing Visible and Thermal IR Signature Data for SSA” 23

24. 24 Anomaly Detection in Non-Imaging