Freeform Object Positioning by 3D Shape Matching Without Artificial Feature Points

 

J.S.M. Vergeest, Y. Song, D. Hartge

Delft University of Technology

Delft, The Netherlands

j.s.m.vergeest@io.tudelft.nl

http://www.dynash.tudelft.nl

 

ABSTRACT

Determining the position and orientation of a workpiece relative to a manufacturing device is a prerequisite for machining path planning. If the workpiece is freeform and if it contains no predefined feature points, then object positioning and tool calibration can be accomplished using shape matching in 3D space. In this paper we study the feasibility and accuracy of non-contact, featureless positioning based on matching of a CAD model to a 3D scan of the object once it is physically mounted. The invariance of the placement of the 3D scanner is verified and the spread in the transformation parameters is analyzed. The matching is performed using a least-squares and a Hausdorff shape difference measure. The effect of point cloud size on the accuracy is analyzed. Recommendations for improvement and applications are provided.