Influence of polarization phenomenology on material discriminability using multi-view polarimetric imagery

摘要

In addition to spectral information acquired by traditional multi/hyperspectral systems, passive electro optical and infrared (EO/IR) polarimetric systems also measure the polarization response of different materials in the scene. Such an imaging modality offers a complete optical description of a surface that can be utilized in identifying objects with complex morphological and camouflaged structures. The polarization property of the radiation from a remotely sensed surface, however, depends on the observation geometry of the system. Therefore, acquiring a polarimetric image in a single viewing direction is not sufficient to improve the material discriminability in the absence of a priori knowledge about the object geometry. Hence, this paper presents a novel multi-view polarimetric system for improving the target-background discriminability. The proposed system takes the approach of imaging the scene in three different viewing directions to infer the physical characteristics of the observed surface by utilizing the angular variation in the polarization response. The sensitivity analysis of the proposed polarimetric system that relates target-background discriminability to various scene related parameters indicates the imaging conditions under which the material discriminability is maximized. Furthermore, scenarios where polarization information can be very useful in improving the target contrast are identified by comparing the detection performance of the proposed system to that of a multispectral system.