From Scar to Scar: Reconstructing Operational Sequences of Lithic Artifacts using Scar-Ridge-Pattern-based Graph Models

Motivated by the importance of lithic artifacts in the development of our ancestors, this study introduces a novel method combining 3D mesh segmentation and graph modelling to determine distinct features of operational sequences being relevant for creating lithic technology analyses.

For analysing the operational sequence of scars, manual scar segmentation was utilized on 3D meshes from both open-access and in-house datasets, to construct directed graph models. These models allow the examination of adjacency and sequential relations among scars, represented as nodes and edges in the graph. Our approach, verified against manual graph models, demonstrates the potential for analysing artifacts digitally and enhancing the understanding of early human technological advancements.

Building on existing practices, we created and approach for determining the relative order of scars. We focus on parameters that are approximations of archaeologically used attributes. For the approximations, we used Multi-Scale Integral Invariants (MSII) curvature values, integral invariants of polylines, and a MSII curvature sampling method along the polylines. In addition, we used parameters derived from either the meshes, like surface area, or the graph models, such as node degree or betweenness centrality. 

The method’s accuracy was tested on various archaeological samples, including Upper Palaeolithic blades and cores from Grotta di Fumane, a Bronze Age blade and an experimental knapping series. Preliminary results for high-resolution 3D models are promising, although the overall accuracy varies. While the results now highlight the need for real ground truth data with low interpretation bias and for improvements of the overall approach, they also demonstrate the potential of graphs modelling operational sequences to predict the temporal relations between adjacent scar. Further, this approach is a step to make the analysis of operational sequences quantifiable, reproducible and reliable beyond the scope of visual comparisons.