A single 3-D feature-based visual servoing

In this paper we present a visual servo control technique by means of a single geometrical feature. The proposed approach refers to a multiple views information and a proper data representation. This work is motivated by tackling the control of an euclidean displacement inside a non-rigid environment, for performing an efficient feature selection and a quick outliers rejection. Precisely, the so-called Plücker coordinates, as originally known to be effective for 3-D straight line modeling, is shown here to be a relevant data representation. With visual measurements, this modeling allows to analyze the visual sensor displacement with respect to the object-of-interest motion, and it is very convenient while designing the state vectors and the kinematic control laws. Two sets of task functions and control laws are presented and these two families of motion are compared according to the considered degrees-of-freedom for rotational and translational velocities. To assess the proposed approach, real experiments with an industrial robot are performed. Results about the cameras displacement show the efficiency of this technique.