This presentation received the "best oral presentation award" at the EARSEL 2009 conference in Tel Aviv, Israel.

1. Adjoint Radiosity Based Algorithms For Retrieving Target Reflectances In Urban Area Shadows Dr. Christoph Borel , Kenneth Ewald, Mark Manzardo, Dr. Charles Wamsley and John Jacobson* Ball Aerospace and *NASIC [email_address] 6 th EARSEL SIG IS workshop2009 Tel Aviv

6. Adjoint Radiosity is uniquely suited for remote sensing ( Inverse Global Illumination Paul Debevec, Siggraph 99 ) Computer graphics rendering: Realistic scene generation through global illumination using known reflectance, geometry and illumination Adjoint radiosity method: Retrieve reflectances from known radiance image, geometry and estimated illumination -> This is the best solution to reflectance retrieval problem! Illumination Reflectance Radiance Image Geometry Radiance Image Illumination Reflectance Geometry Remote sensing problem Computer graphics rendering problem unknown known Forward problem Inverse problem

7. New method needed to perform reflectance retrieval Note: Because of complexity of problem our current focus is limited to illumination correction Atmospheric Correction Calibration HSI Sensor DN L m (x,y, λ ) L sky ( θ , φ , λ ) Current focus Illumination Correction ρ (x,y, λ ) 3-D model Sun L surf (x,y, λ ) Atmosphere

8. 2-D corner radiosity model with direct solar and sky light 1 2 θ sun F 12 >>F 21 S=tan( θ sun ) B 2a B 2b B 1 f 1

10. Fast skyview factor using shadow volume Shadow volume (SV) of DEM DEM of London Shadow=(DEM eq SV) Add all shadow images Reference: Ratti C, Richens P.,1997 http://www.bath.ac.uk/pip/directory/showperson?personId=502162 Principle: A surface is in shade if the height of the shadow is higher than the surface.

11. A walk in downtown Dayton using LIDAR data

12. Experiment to retrieve illumination