Bio-inspired object classification using polarization imaging

Abstract

Polarization is an inherent property of light. The phenomenon of polarization by reflection of light from a transparent object differs from an opaque object. The specular reflection from the transparent object is highly polarized compared to the diffuse reflection from the opaque object. The differences in the polarization pattern can be recorded and can be used in machine vision applications like object classification and autonomous agent navigation. In this paper we present different methodologies like degree of polarization, polarization Fresnel ratio, Stokes degree of polarization to classify among transparent and opaque objects. Based on the polarization profile, the shapes of the transparent objects are also estimated. © 2012 IEEE.