Neural network-based image moments for visual servoing of planar objects

In this paper, Neural Network (NN)-based image moments are proposed to address the challenges of choosing proper image features for Image-based Visual Servoing (IBVS). The proposed two NN-based image features can estimate the rotational angles around x and y axes of camera frame for planar objects. Based on the proposed image features and other 4 commonly used image moments, the interaction matrix relating the chosen image features to camera motion is derived to have maximal decoupled structure. In addition, the analytical form of depth computation is given by using classical geometrical primitives and image moment invariant. A proportional IBVS controller is designed based on the derived interaction matrix and the tracking performance is thus enhanced for the 6 degree-of-freedom robotic system. Simulation results on a 6-DOF robot system are provided to illustrate the effectiveness of the proposed method.