Adaptive visual servoing using curve features with unknown geometrical parameters

This paper presents a novel curve image feature description using cubic Bezier curve, which extends the depth-independent image Jacobian matrix to curve features. We derive the image Jacobian matrix that can be linearly parameterized by the unknown curve feature geometrical parameters in the 3-D space. To estimate the unknown parameters online, we propose an adaptive algorithm using the defined curve image features. The control law is designed and the asymptotic stability is analyzed by the Lyapunov theory. There is no need to known the curve parameters or do correspondence matching while visual servoing. Experiment has been conducted to demonstrate the performance of the proposed controller with the curve features.