Research On Robotic Visual Servo System

This paper is based on the image processing algorithm and sliding mode variablestructure control theory, and choose mechanical arm and ping-pong as the research object,working out a software programme for image segmentation technology and servo controlsystem by VC. To realize the robot has able to grab target according to visual information,which is need to extract features of the target from2d image from camera, but the imagewas disturbed from the uneven light, leading to outline of the image appear rupturephenomenon, which influence recognition effect of the target feature, finally, this will leadto poor performance of visual servo system in the real environment. Pointing to theproblem, filtering methods have been input to remove interference and try to use thesupport vector machine method to do image processing in this paper.SVM (support vector machine) has able to classify nonlinear image data effectively,which is very suitable to segment target image of the robot visual system. This paper usingsupport vector machine classify image data of the target, the experimental results show thatif only choose appropriate sample size and feature, this method has able to segment targetimage with fuzzy boundaries and gray uneven, and overcome uncertainty problems thatrely on extensive experiments to determine segmentation thresholds.As for the modeling error and uncertain problems in the manipulator, this papercombines the sliding mode variable structure control method that often application forrobot, and analyzes chattering problem which existing in the variable structure, and carryon simulation research to trajectory of the manipulator. The paper extended it into a slidingmode variable structure control method based on the radial Basis Function Neural Networkand the name model to weaken chattering when sliding mode control manipulatormovement and analysis the stability of the manipulator movement by the Lyapunovfunction, and the simulation results verify that the method is feasible.