Transcript Slide 1
Polarimetric Imaging Sensors for Surveillance Navigation and Communications Howard Schultz and Andrés Corrada-Emmanuel University of Massachusetts, Aerial Imaging and Remote Sensing Laboratory [email protected], [email protected] Chris Zappa and Michael Banner Columbia University, Lamont-Doherty Earth Observatory [email protected], [email protected] Office of Naval Research January 3, 2007 Long Term Goals • Develop passive remote sensing techniques for studying the dynamics of the upper ocean • View the surface environment from a submerged platform – Polarimetric Periscope – Uplink/downlink Communications Optical Flattening Motivation • View the above-surface environment from below the surface • Objects in a scene taken from underwater are naturally blurred by wave motion – Image sharpening – wave estimation algorithm • Wave estimates are not yet accurate enough to substantially improve the reconstructed images above the surface beyond what can be achieved assuming a flat surface. Optical Flattening • Use information about the 2D slope field of the ocean surface to remove image distortion -- What would an image taken through the ocean surface look like if there were no waves? • Real-time processing Projective Image Formation Model Air Observation Rays Water Imaging Array Exposure Center Optical Flattening Algorithm* • Collect polarimetric images • Recover the 2D surface slope field • Compute the refraction for each rays as it passes through the air-sea interface • Create an undistorted image (sort on the direction of the rays in air) *Patent Pending Process, University of Massachusetts, Amherst Degree of Linear Polarization (DoLP) vs. Incidence angle 1.0 Reflection 0.8 0.6 Q U DoLP I 0.4 2 2 Refraction 0.2 0.0 0 10 20 30 40 50 60 70 Incidence Angle (degrees) 80 90 Ray tracing image formation model A lens maps incidence angle θ to image position X θ Lens Imaging Array X Ray tracing image formation model A lens maps incidence angle θ to image position X θ Lens Imaging Array X Ray tracing image formation model A lens maps incidence angle θ to image position X Lens Imaging Array X Ray tracing image formation model A lens maps incidence angle θ to image position X θ Lens Imaging Array X Ray tracing image formation model A lens maps incidence angle θ to image position X θ Lens Imaging Array X Refraction Air Water Distorted Image Point Refraction Air Water Distorted Image Point Undistortion Compensating for Refraction Air Air Water Distorted Image Point Undistorted Image Point Undistortion Compensating for Refraction Air Air Water Distorted Image Point Undistorted Image Point Implementation Considerations • Uses only one polarimetric camera • Exploit the natural time scale separation tsky > tobjjects > twaves > tshutter to estimate the polarization distribution of the sky radiance • Real-time requires a functional approximation between the inferred incoming Stokes vector, the observed scattered Stokes vector and surface slope (Kattawar, 1994; Voss and Fry, 1984; Sabbah and Shashar, 2006; current effort in RaDyO). • Statistical techniques will always be needed to sharpen final image. • Requires a precise motion package