3D Location of Circular Features for Robotic Tasks

In many robotic applications, 3D location estimation of an object with respect to a reference frame is required. Circular markers located on the object are often employed for this purpose. This paper addresses the problem of 3D location estimation of circular features from a single camera image. A mathematical closed-form solution of the problem allows to determine the normal vector to the circle feature with respect to the camera reference frame. High calculation accuracy is guaranteed by a Kalman filter which performs a sub-pixel parametric estimation. A set of experimental results shows the validity and the accuracy of the process involved in 3D angular estimation. The theory is applied to a real industrial problem, namely the 3D location of a car rim. In fact, this information is necessary to perform a pick-and-place robotic task.