Abstract
Due to the mass of sherds and vessels that have been dug out by archaeologists over many years, the request for an automatic classification system has become obvious. This paper presents algorithmic key ingredients of such a system: segmentation, feature extraction and clustering. Therefore, the extensive analysis of a representative profile line is of high importance, because it helps not only to retrieve global measures and a segmentation of the body, but also to separate attachments like handles, lugs, feet or spouts for gaining an even more detailed description of the vessels. Finally, we present ways to use the extracted features for queries and hierarchical classification.