Eleonora Marconi, Antonio Budano, Giancarlo Della Ventura, Federico Fina, Alberto Botti, Sandro Tassa, Ottavia Palacino, Lorenzo Conte, Marianna Franco, Francesco Pacetti, Caterina Coletti, Armida Sodo, Luca Tortora
AI-assisted Reconstruction of Archaeological Pottery from digital 3D mesh models
Ancient pottery in archaeological sites is typically found as broken fragments. The collection, classification, and assembling of those pieces into their original artifact may take years of hard work, especially when the fragments are irregular, intermixed with parts of different vessels, or if some key pieces are missing. This problem is traditionally handled via two main steps: (1) the Classification of Archaeological Fragments into similar groups (CAF) and (2) the Reconstruction of each group into the original Archaeological Objects (RAO). Over the years, many alternatives have been proposed to solve this problem. A seminal approach was exploiting the color and texture properties of the fragments. More recently, the use of 3D computer-aided reconstruction methods gained attention as promising tools in pattern recognition. For this reason, researchers have implemented algorithms to collect all the information necessary to reconstruct a complete vessel from suitable data collected via 3D scanners. In this work, four types of algorithms were tested to reconstruct the objects without an a priori knowledge of the final shapes. The method exploited the geometric features obtained from the 3D mesh model acquisition on artificial samples from a broken mug, used as test cases. The best algorithm satisfying the final 3D reconstruction was then applied to the study of archaeological ceramic fragments from Villa della Piscina in the Parco Archeologico of Centocelle (Rome, Italy) within the project ERCOLE. The aim of this work is at developing a tool that satisfies the criteria of accuracy, performance, robustness, transportability, cost, and careful handling of archaeological specimens.