Design Of Visual Servo Control System For Robot Based On BP Neural Network

People get more than 80% information through eyes. If we add vision sensor to robot, it will be more agile. Hence such problems as teaching and playback mode of current robot system and poor flexibility will be resolved. Research of visual servo robot has been becoming hot in the domain of robot recently. Because some kinds of deviations will be produced in the process of robot manufacture and application, and lens distortion is also inevitable under some conditions, highly precise calibration of robot and camera is needed. However, calibration is a kind of time-consuming and laboursome work, furthermore it correlates closely to production circumstance. Therefore many scholars have commenced studying visual servo system for robot without calibration. With the development of Artificial Neural Network (ANN) technology, ANN has been widely used on sundry occasions. As a result of its complex nonlinear mapping capacity, ANN is very suitable for visual servo system for robot without calibration.In this dissertation, a visual servo system for robot based on a SCARA teaching robot was constructed, and the control mode of visual servo controller was based on back propagation (BP) neural network technology. In this paper position information of robot joints was introduced into visual servo controller, so efficiency and precision of controller were increased significantly adopting"stepwise process in small scale"method. This paper made a deep research on image processing of computer vision. We studied target identification and extraction algorithm based on color image and grey image respectively. Towards round target, we adopted fitting method of round based on least square principle to achieve localization of gravity center, and its positioning accuracy will be improved. With the characteristic of servo control for SCARA robot, motion elements of robot adopted path planning based on continuous path (CP) to control the end trace of robot. The visual servo system for robot could complete localization of arbitrary target on working platform with high precision. And finally, the validity of control mode was verified through experiments.Above all, this paper made an intensive study of visual servo system for robot based on neural network, adopting some new processing methods and control ideas. So it is expect to have some positive influence on succeeding researches.