Method of recognising the two-dimensional position and orientation of already known objects.

摘要

Pattern recognition in the industrial sector employs for the most part three-dimensional image analysing processes. In order to carry out simple, especially planar recognition tasks with respect to the position and orientation of objects (1) whose form is known, recourse is made likewise to three-dimensional recorded image data. Due to the prevailing limitation of the degrees of freedom, however, it is necessary to treat as variables only two coordinates of a plane and one angle of rotation phi , which characterises a rotation of the known object (1) inside of the plane. The intersection with a primitive volume (2), especially a circular cylinder, yields lines of intersection (3), which are to be employed as comparative criteria between a measurement object and a reference position. The lines of intersection (3) are constant in shape and are described only as a function of the angle of rotation phi .