Similarity Analysis of Archaeological Potsherds Using 3D Surfaces

This work presents a new methodology for efficient scanning and analysis of 3D shapes representing archaeological potsherds, which is based on single-view 3D scanning. More precisely, this work presents an analysis of visual diagnostic features that help classifying potsherds, and that can be automatically identified by the appropriate combination of the 3D models resulting from the proposed scanning technique, with robust local descriptors and a bag representation approach, thus providing with enough information to achieve competitive performance in the task of automatic classification of potsherd. Also, this technique allows to perform other types of analysis such as similarity analysis.