Positioning a camera with respect to planar objects of unknown shape by coupling 2-D visual servoing and 3-D estimations

This paper proposes a way to achieve positioning tasks by 2-D visual servoing when the desired image of the observed object cannot be precisely described. The object is assumed to be planar and motionless, but no knowledge about its shape or pose is required. First, we treat the case of a threadlike object, and then we show how our approach can be generalized to an object with three particular points. The control law is based on the use of 2-D visual servoing, and on an estimation of two 3-D parameters. We show that this control scheme is not sensitive to the calibration of the camera. We conclude this paper by experimental results relative to objects of unknown shape. In addition, an algorithm to estimate the depth between the object and the camera is provided, which finally leads to a 3-D estimation of the object shape.