Polarimetric Modeling and Visualization

Polarization is a fundamental property of electromagnetic radiation and has utility at EO-IR wavelengths. Therefore, polarization can be exploited to suppress natural clutter, improve detection performance of man-made target objects, provide shape and contrast information, and compensate atmospheric scattering. This proposal will develop a modeling framework for the accurate prediction of polarimetric signatures based upon a simplified set of parameters to determine polarization state. This proposal will also develop novel display strategies that intuitively present otherwise complicated polarimetry information. An accurate and straightforward modelling tool compatible with human cognitive analysis will broaden the user base of polarimetry data. Reflective polarization phenomenology at EO-IR wavelengths is complicated to model due to dependency on material shape and chemistry, sensor-illumination geometry, scene temperatures, and atmospherics. These models require full material characterization in the form of the pBRDF, however pBRDF measurements are difficult to obtain for many remote sensing targets of interest because of indoor laboratory constraints and outdoor variables such as atmospheric conditions, source illumination position, and source polarization state. Consequently, pBRDF measurements are often not available for materials of interest at the level of fidelity required by existing models. Existing modelling tools, such as NEF and DIRSIG, are powerful and based upon validated polarimetric BRDF models, but they are limited in their applicability to materials that have been fully characterized within their databases. Moreover, the these tools require a significant amount of domain-specific knowledge to use and interpret the produced outputs. This proposal will create simplified models that can predict expected polarimetric signatures without having fully characterized pBDRF models of each target material. The simplified models will be based on empirical data measurements while using high fidelity theoretical models for validation. Legacy approaches for visualizing polarimetric imagery is a monochromatic display of the Stokes vector and derivative DoLP and AoP images, or mapping the Stokes images into RGB or HSV. A user with limited familiarity of polarization find exisitng display strategies non-intuitive and can confusing. For example, display strategies that can enhance an S0 image with polarimetric information to improve target contrast or shape information,  while keeping polarization complexities transparent to the end user, would be valuable to an imagery analyst. We will build upon these concepts to develop sophisticated and novel polarimetric display strategies to offload the human cognitive load required to convert complicated polarization information into readily actionable intelligence. The developed strategies will also be developed with the goal of providing improved object detectability for automated detection algorithms.