Constant radius geometric features segmentation in archeological pottery

This paper gives a contribution to the automatic recognition of significant features of ancient ceramics, useful for the historical and/or archeological investigation. These very common type geometric features are obtained by a sweeping action that leaves negative or positive traces, characterized by a cross section with one or more constant radii. The paper proposes a novel methodology that, analyzing the principal curvatures at the points of high-density geometric models of ceramic vessels, acquired by laser scanning, identifies the nodes potentially attributable to these features of constant radius. The recognition process is not trivial since it is affected by uncertainties. To overcome the limits of a recognition based on crisp sets, the recognition rule, proposed for the feature segmentation, is implemented by a fuzzy approach. The method has been tested in the identification of embossed decorations in an ancient olla and it proves to be promising for further applications on other types of geometric features of constant radius.