A simple adaptive control for visual servoing

Abstract

Traditionally, visual servoing system is divide in position-based visual servoing system (PBVS) and image-based visual servoing system (IBVS). PBVS needs to estimate the pose of target in 3-D space, and IBVS often control the feature in 2-D image plane. The performances of both systems depend on the selection of the state vector. So- it is important to select a desirable state vector to design a visual servoing system. Analyzing the affine model of image motion, we find when camera motion consists of only translation motion and the target is one rigid planar surface, the projected point position of the target centroid in image plane --[x/sub c/,y/sub c/]/sup T/, and the radio of two depths--s, can explicitly describe the target pose with respect to the camera. In addition, x/sub c/,y/sub c/, s can be easily calculated from image moments which are quickly computed from a binary image. So, in this paper, the vector [x/sub c/,y/sub c/,s]/sup T/ is used as the state vector. Then a simple image Jacobian matrix is gotten from the state equation and leads to a simple adaptive controller, which can drive a camera to the ideal position.