Decoupled control for visual servoing with SVM-based virtual moments

Abstract

In order to improve the performance of image based visual servoing (IBVS) system, a novel moments based method with good decoupling and linearizing properties is proposed. Two virtual moments are designed to decouple the rotational motions around x-axis and y-axis of camera frame. Support vector machine regression method is proposed to learn the mapping model from four moment invariants to the virtual moments. With ESM and resolved-rate methods, the robotics IBVS controller is designed combing virtual moments and other four classical moment-combination features. Simulation results are provided to illustrate the effectiveness and optimal performance of the virtual moments and the proposed robotics IBVS controller.