Omnidirectional Egomotion Estimation from Adapted Motion Field

Abstract

The egomotion estimation problem is basic task to most of vision-based mobile robot applications. Recent research has shown that the use of omnidirectional systems with large field of view facilitates the computation of the observer motion. All previous work takes a motion field, computed in the image, as an input in egomotion estimation process. This motion field is calculated using classical algorithm used for perspectives images. In this paper we present a method for the recovering of egomotion using an adapted motion field. Based on a motion model and adapted neighborhood described for parabolic mirror, this adapted motion field allows succeeded estimation of egomotion in paracatadioptic images. Experimental results are shown and comparison of error measures are given to show that the use of an adapted motion fields improve the estimation of the observer motion.