58 — Identifying Designs from Incomplete, Fragmented Cultural Heritage Objects by Curve-Pattern Matching

Many cultural heritage artifacts feature large, embellished, curved designs (that are often fascinatingly similar to space-filling curves). Because these artifacts are often in the form of shattered pottery (“sherds” — which I put in quotes here because this is the first time I’ve ever seen that word, somehow), these curves are difficult to reconstruct. In addition, the 3D structure of the original pottery may be impossible to reconstruct based on existing fragments; but by using curve-matching algorithms such as those described here, both pattern design reconstruction and structural 3D reconstruction may be possible.

Chamfer matching — an algorithm that uses unique features of an image to determine colocalization — isn’t feasible here because so often the designs being matched are ‘stamped’ or duplicated across the surface of pottery.

This paper focuses on the paddle-stamping tradition from south-east North Native American pottery (c. AD 500). Paddles were designed with stylized curves, and variations in these curves can be used to trace lineages of this paddle tradition between and across communities.

Because these curve designs changed gradually over generations, known databases of existing designs can be used as a matching-corpus for newly acquired sherds. These new patterns can then be scale-dependently matched and added to the corpus.

The software pipeline is currently all MATLAB, as far as I can see; I think there are some great strides that can be made by improving the curve-detection algorithm (in steps 3-6 in Fig 5) to go from image to a set of higher-fidelity curves. It also looks like there is not a great deal of 3D unwrapping taking place; I wonder if more matches would be found — or matches with greater confidence and overlap — if the slight curvature were taken into consideration.

This matching system appears to work competitively with existing solutions for partial-to-global curve-matching, which is exciting — this means that much larger sets of these paddle curve designs can be reconstructed and interpreted; and I imagine that this matching system can be extended to many far-reaching applications both inside the field of archaeology and out.