Visual Servo Control Of Underwater Vehicles Based On Image Moments

Traditional image-based visual servo control methods of underwater vehicles directly use point features in the image plane.In these controllers,image Jacobian matrix of the system is heavily coupled with strong non-linearities,and the local minimum exists due to the redundant point features,which makes the system sensitive to image noises and uncertainties.In this thesis,hybrid image moments visual servo control,attitude estimation visual servo control,image jacobian matrix estimation are studied for underwater vehicles.The main contributions are as follows.1.Hybrid visual servo control of underwater vehicles based on three image moments and four image moments.The hybrid visual servo model using three image moments and four image moments for underwater vehicles is designd by combining the normalized centroid image coordinates,normalized area,heading angle and three Euler angles of underwater vehicles.The theoretical analysis and simulation experiments prove that the three image moments based hybrid visual servo system is applicable to arbitrary shape objects,and has decoupling characteristics with the motion of translation DOFs,thus the three-dimensional motion trajectories are optimized.the four image moment hybrid visual servo system can solve the problem that yaw angle can not be measured,but the system is only applicable to the target object shape with different lengths of long axis and short axis.2.Image-based visual servo control of underwater vehicles with attitude estimation.To solve the problem that the attitude angle of underwater vehicles relative to the plane of the visual target can not be measured,and to overcome the coupling of the pitch and roll angles,the inadaptability to partially shaped objects in the traditional methods,a neural network is used to estimate the pitch and roll angles of underwater vehicles in the image visual servo controller.The simulation results show that the attitude angle estimation based on neural network and the visual servo control of underwater vehicle image are all effective.3.Hybrid visual servo control based on image depth estimation and image Jacobian matrix estimation.Two methods are used to estimate the image Jacobian matrix.1)When the camera parameters are known,the image depth is estimated by the relationship between the zero-order image moment and the image depth,so that the image Jacobian matrix can be calculated online.2)The dynamic Broyden method is used to obtain the numerical solution of image Jacobian matrix online.The simulation results show that these two methods can estimate the image Jacobian matrix online for the visual servo control of underwater vehicles,and these two methods make the system converge faster,more stable and more accurate than the visual servo control with constant approximation.