Object pose determination from range data

This paper is concerned with determining the position and orientation of a three-dimensional object from range data. An algorithm is presented which uses an enclosing object and the central moments for an initial estimate of pose. A refined estimation is obtained by a nonlinear least squares technique. The use of an enclosing object reduces the number of parameters which must be solved for by the nonlinear least squaes estimation. An object representation based on a relational graph of superquadric primitives is used for description so that the method is applicable for 3-D objects of general shapes.