A self-calibrating technique for visual servoing

Abstract

In this paper, a self-calibration procedure which is carried out in parallel to visual servoing algorithms is proposed. Our approach exploits the singular values property of the essential matrix and it is based on the introduction of a suitable cost function and a criterion for minimizing it on-line in order to update the current estimate of the intrinsic parameters. The behaviour of the visual servoing is hence improved for the case of unknown intrinsic parameters since the self-calibration procedure allows the robot control to work closer to the ideal case. Simulation results for position-based, image-based and 2 1/2 D visual servoing show also that the correct estimation of the intrinsic parameters is achieved in few steps of the required ones by the visual servoing algorithm to converge.