Precise pose and radius estimation of circular target based on binocular vision

Abstract

A novel method of estimating circle pose and radius is proposed, based on binocular vision. This method uses the special vectors that are irrelevant to the unknown radius to calculate the normal vector of the circular plane, which is the foundation for calculating the vanishing line of the circular plane in the imaging plane. The projections of the circle center on the left and right image are calculated afterwards through the vanishing line. Then, the circle center is reconstructed by the triangulation technology of binocular stereo vision. Finally, the radius of the circular target is calculated through the relative positions of points in the same coordinate system. The estimates of normal vector, center position, and radius size are unique and definite, so there is no need for a subsequent screening process. When the radius is unknown, most of the conventional circle pose estimation methods are unusable. However, the proposed method has no prerequisites about the circular target in order to estimate the pose and radius of the circular target, which improves flexibility and extends the application field. Besides estimating the pose of a circular target of unknown radius, the proposed method can also be applied to the vision measurement of the radius of a circular object. Both actual measurement experiments and noise simulation experiments are performed. The results show that the proposed method is more precise and robust than conventional methods.