LOCATING OVERLAPPING INDUSTRIAL PARTS FOR ROBOTIC ASSEMBLY

The problem of recognising overlapping parts is of considerable intere st in industrial automation. While it Is possible to employ carriers o r pallets to separate or prearrange the parts for robotic assembly tas ks, a vision system which can recognise the parts even though they may be partially overlapped and in random positions is much flexible. In this paper, we employ a Hough transform technique to determine the pos es (translations and rotations) of overlapping parts in the image. Sha pes of parts are broken down into fragments, each representing either a line or a circular arc primitive. The geometrical relations and prop erties of primitives are used to derive the matching of primitive pair s between the model and the image. This paper also presents methods to determine the top-bottom relation of overlapping parts in grey-level image by tracing the shadowed points along the overlapping boundaries. The detected poses along with the top-bottom relation of overlapping parts provide sufficient information for robotic assembly in complex e nvironment.