We discuss an implemented software system that interprets dense range images obtained from scenes of heaps of postal pieces: letter, parcels, etc. We describe a model-based system consisting of segmentation, modeling, and classification procedures. First, the range image is segmented into regions and reasoning is done about the physical support of these regions. Second, for each region several possible 3-D interpretations are made based on the partial knowledge which is updated whenever a new interpretation is obtained. Finally each interpretation is tested against the data for its consistency. We have chosen the superquadric model as our 3-D shape descriptor, plus deformations such as tapering and bending along the major axis. The superquadric model is an analytic representation of volume for which a cross-section is one of a class of curves varying between rectangular to elliptical shaped. Superquadric parameters are recovered by minimizing the least-squares error between the superquadric surface and the range data. The system recovers p position, orientation, shape, size and class of the object. Using the goodness of fit and Euclidean distance measures, and the shape and size parameters of the recovered model, objects are classified into one of the following broad categories: flat (letters), box (parcels), roll (circular and elliptical cylinders), and irregular (film mailers etc.). The overall approach to this problem has been to find the most general yet computationally economical method to interpret the data. Experimental results obtained from a large number of complex range images are reported. Comments University of Pennsylvania Department of Computer and Information Science Technical Report No. MSCIS-90-14. Author(s) Ruzena Bajcsy, Kwangyoen Wohn, Franc Solina, Alok Gupta, Gareth Funka-Lea, Celina Imielinska, Pramath Sinha, and Constantine Tsikos This technical report is available at ScholarlyCommons: https://repository.upenn.edu/cis_reports/535