Passive 3D Reconstruction using Polarimetric Infrared Imagery

Abstract

Passive 3D sensing and reconstruction is an important area in computer vision, robotics, surveillance, reconnaissance and automatic target recognition fields. In the past methods such as stereo, shape from shading, and reflectance (visible band) polarization have been tried. Some of more obvious advantages of passive 3D sensing and reconstruction are its reliance on relatively low-cost sensors its non-intrusive properties. This paper summarizes methods for extracting 3D information and index of refraction of the emitting surfaces in scene by means of a pair of polarimetric passive infrared imaging sensors. These sensors operate mainly in the thermal region of electromagnetic spectrum. Each sensor provides the Stokes vector at each sensor pixel location, from which, degree and angle of linear polarization are computed. Angle of linear polarization provides the azimuth angle of the surface normal vector at each pixel location. The depression angle of this surface normal, however, depends on the degree of linear polarization and index of refraction of the viewed surface. We present two methods to estimate the depression angles of the surface normals in a scene without relying on the a priori knowledge of indices of refraction. In one case a set of distributed sensors are used. These two are placed in geographically distinct positions but are otherwise identical. In the 2nd case the two sensors are co-located but have different spectral bands.