Polarization-Based Orientation in a Natural Environment

Many insects, birds and aquatic animals are sensitive to the polarization state of light and are able to use this information for orientation in their habitat. In this paper we imitate this ability and evaluate the possibility of a computer vision system using a standard CCD-camera to determine directions in outdoor scenes based on the polarization pattern of the sky. Even if the sky is covered with thick clouds, the incident light has a sufficient degree of polarization. A physics-based model is proposed to describe the state of polarization of the skylight, including the depolarizing effect of clouds and dust in the atmophere. Based on our polarization model we propose features for orientation that are invariant to weather and other environmental conditions. Experiments show that through polarization analysis we can get the same precision in determing a direction as by using a compass.