Tracking features with camera maneuvering for vision-based navigation

In this paper we present a method to obtain corresponding contour features in a sequence of images. Our final goal is to compute camera motion and structure, for navigation based on vision. We make a feature tracking in the image considering not only the contour geometric information, but also a brightness-based description. A kinematic model of brightness attributes has been proposed in order to obtain a brightness evolution constraint more reliable than the classical brightness constancy model. Camera maneuvers, like sudden orientation changes, are detected, estimated and used to obtain a better correspondence of features. Experimental results, using real images, are shown. Our main contributions are the proposed evolution model of brightness attributes and the detection of camera maneuvers.