Robust matching area selection for terrain matching using level set method

Abstract

To enhance the reliability of path planning in scenery guidance system, it's very important to select reliable or high matching probability areas from the navigation reference images for performing unmanned aerial vehicles localization. This paper applies three measures and proposes a new selection scheme base on a simplified Mumford-Shah model. The proposed method artfully avoids selecting thresholds to separate the feature images and optimally selects robust-matching areas by evolving the level set function. Experiments of the selection show that the proposed method is efficient. © Springer-Verlag Berlin Heidelberg 2005.