A Unified Initialization Phase to Improve Visual Servoing in an Unknown Environment

In this paper, we consider the problem of the accurate estimation of the depth in the specific context of 2D visual servoing. To this aim, we have developed a predictor/estimator pair. This depth will then be used to solve simultaneously three problems, namely the determination of the interaction matrix, the estimation of the visual features (when the image is temporarily unavailable) and the automatic computation of their reference values. Simulation results validate our work.