The Research On Robust Vision Servoing Control System Based On The Dynamic Image Sequence

It's an important developing direction that the robot is accompanied by vision which also called vision servoing control system. To improve the accuracy and range of application, image preprocessing, object tracking, camera calibration and designing of visual servoing controller are deeply researched. About of all, the problem of noise-reduction, enhancement, and edge detection and so on are very important to image preprocessing. There are two methods developed, one is a method of image enhancement based on homomorphic filter and, the other is a new thresholding method for edge detection based on Radon Transform. Then, in order to resolve the tracking problem in the dynamic image sequence. A simple approximation correlation tracking algorithm for series Images is proposed. If the motion between two frames is approximate, the method could reach the desired position quickly under certain rules. The only regret is accuracy. To improve it, another algorithm is introduced, based on the image moment invariant. In our research, it only need to calculate four first order image moment invariant to obtain the ideal object feature, the experiment result shows the validity of the proposed algorithm. Following that, there are two visual servoing algorithm has been described. One of method is robust adaptive feedback control algorithm, which treat the object tracking as a pre-estimate problem of nonlinear optimal control, employs the extended Kalman filter combined with recursive least square, to find a balance between calculate cost and performance index, so that the system in an optimal situation during the tracking processing. The other algorithm which is based on the Image Jacobian Matrix, we utilized the extended Kalman filter to pre-estimate the system state and get real-time parameter, obtain desired controlled variable, strengthen ability of object tracking under a given range. Finally, an improved method of camera calibration has been developed. According to the characteristic of the target, here only study the single camera calibration. Based on the famous algorithm, Two-stage of Tsai, Incorporating with traditional linear camera model, a new camera calibration method is given. Compared with classical method, our method seemed easier to calibrate with proper parameter, has better robust.