Automatic Shape Feature Recognition for Ceramic Finds

Abstract

Ceramic sherds are the most common finds in archaeology. They are complex to analyze and onerous to process. A large number of indistinct sherds coming from excavations must be preliminarily grouped in some categories. This clusterization helps the next phase, in which archaeologists classify the ceramics. Due to the difficulty of these preliminary, repetitive, and routine phases, a great deal of archaeological material remains unstudied in museum repositories or archaeological sites. An effective method to automate these routine phases is presented in this article. The proposed method performs a shape feature segmentation of the sherds, which is fundamental to undertake any further analysis, such as potsherds classification, reconstruction, or cataloging. A set of specific shape features, useful to understand the find properties, is defined and methods for recognizing them are proposed. The method's performance is tested in the analysis of some real, critical cases.