Depth estimation to manage visual signal loss during visual servoing with a 3 DOF camera

This paper deals with the problem of estimating the visual features during a vision-based navigation task when a temporary total occlusion occurs. The proposed approach relies on an existent specific algorithm. However, to be efficient, this algorithm requires highly precise initial values for both the image features and their depth. Thus, our objective is to design a predictor/estimator pair able to provide an accurate estimation of the depth value, even when the visual data are noisy. We also aim at obtaining a method reducing the implementation complexity while preserving performances. The obtained results show the efficiency and the interest of our technique.