Recovery of mail piece shape from range images using 3-D deformable models

Automatic sorting of all classes of mail must take into account different shapes and sizes of mail pieces. A system for classification of mail pieces according to their shape using range images is proposed. Volumatric models of single mail pieces characterized by position, orientation, size and shape parameters are recovered using least squares minimization of a model fitting function. The models used are superquatrics with global deformations. The recovered parameters can serve for different classification schemes which depend on the particular methods used for the manipulation of the mail stream. A classification scheme that divides mail pieces is presented.