Determining pose in a multiple-view based 3-D object recognition system

The authors address the pose determination problem in constructing multiple-view-based three-dimensional (3-D) object recognition systems. They propose a procedure which performs a sequence of 3-D transformations on the model object so that the pose of the unknown object can be determined. A series of 3-D transformation matrices is derived. The composition of these matrices relates the unknown object to its corresponding model object. A procedure to determine the pose of the unknown object by combining the derived transformation matrices is described. The experimental results that corroborate the proposed method are reported