We describe a method to achieve robotic positioning tasks by image-based visual servoing when the object being observed has a complex and unknown shape. We first focus on the computation of an analytical expression of the interaction matrix according to a polar description of the image contour of the object. Experimental results are presented to validate the proposed algorithm. In particular, the robustness of the control law is tested with respect to a coarse calibrated system, to an approximation of the depth of the object, and to partial occlusion.
展开▼
