Aiming at problems that in eye-in-hand visual servo system

traditional Jacobian matrix is difficult to determine the parameters and uncertainty of target motion

a visual servo method for on-line estimation of composite image Jacobian matrix is proposed. The Kalman filter is utilized for online estimation of composite image Jacobian matrix

and a visual servo controller is designed based on the Lyapunov principle

it can realize the asymptotic convergence of the image characteristic error and the stable tracking of the moving target. The simulation of planar two-link robot validates the effectiveness of the algorithm.