Preprocessing of 3D-Data for Classification of Archaelogical Fragments in an Automated System

Thousands of fragments of ceramics (called sherds for short) are found at archaeolgical excavation sites. Till today archaeologists have drawn and classified them manually. This method is very timeconsuming and classification depends on the expierences of the archaeologists. So we developed a system for classification based on 3D-models of the sherds acquired by a 3D-scanner using stuctured light. The 3D-model is properly orientated by estimating the rotational by the use of plane-fitting and a Hough-inspired method. Afterwards the longest elongation (called profile) around the wall of the sherd is calculated. The extremal points of the curvature of the profile are calculated to segment the curvature into its primitives. The classification is done by comparison of the extremal points and primitives as it is done by archaelogists. Experiments show that the automated system provides a fast and consequential classification.