Analysis of Hyperspectral Data Using Polarimetric Characteristics

摘要

The utility of polarimetric reflectance characteristics of targets and backgroundsurfaces in the analysis of hyperspectral imagery data is investigated. A technique is proposed for filtering a data hypercube of an imaged scene to select targets for subsequent analysis using standard hyperspectral signature matching techniques, thereby reducing image analysis time. An experimental study to measure polarization characteristics of reflected light from various surfaces in order to determine wavelengths for maximum and minimum intensity differences between polarized reflectance values is proposed. A second study is outlined for collection of simulated hyperspectral imagery that would attempt to validate the proposed filtering technique. A research of past studies indicate that useful polarization signature components are present for many targets and target materials. Additionally, backgrounds composed of grass, trees, dirt and clouds generate very little polarized components making detection of targets using polarization signatures feasible.