Visual Servo Control of Underwater Vehicles Based on Image Moments

Abstract

To solve the dynamic positioning problem of underwater vehicles for executing autonomous operation tasks, an image moments-based six degrees of freedom (DOF) visual servo control method is proposed. At first, the equations of motion of underwater vehicles are presented, and the image moments of underwater objects are introduced. Then the Jacobian matrix of image moments is derived, and the image- based visual servo control algorithm is designed, in which the feedback states are constructed by the image moments and attitude angles of the vehicle. To estimate the pitch and roll angles, a multi-layer neural network is trained to approximate the angles with image moments. The stability of the proposed visual servo control is analyzed by a Lyapunov-based method. The simulation results prove that the proposed control method has satisfactory performances for decoupled control of different DOFs with underwater targets with different shapes.