A new Approach for Structure from motion Underwater Pile-Field Documentation

For a pilot study carried out by the University of Bern together with local partners in Summer 2018 at the pile-dwelling site Bay of Bones (Rep. of Macedonia), a new workflow for underwater pile-field documentation was developed.

The site lies in shallow water of 3–5 meters depth and the most obvious constructive remains of the prehistoric settlement are thousands of wooden piles. The piles, mainly of oak and juniper, are excellently preserved in the lake sediments. The aim of the project was to document and sample 40 m² surface area of the pile-field and the dendrochronological analysis of the samples.

Dendrochronological sampling requires cutting the top-ends of the piles and thus changes the preserved situation. Therefore beforehand documentation must ensure the localization of each pile on a map.

This calls for a method that ensures a) that every pile is distinctly labeled and b) the location of each pile is accurately captured. While on land, this can easily be achieved, underwater working conditions complicate common procedures. E.g. by measuring with a folding ruler from a local grid, there is later no way to evaluate measuring mistakes or the internal error of the local grid. In addition, for unpracticed divers measuring by hand underwater is not only time-consuming but also tends a lot more to erroneous results than on land.

The goal was therefore to find a time-saving, accurate and easy to carry out way to locate the positions of several hundred piles in shallow water. The best solution for us to achieve these goals was a new standardized and reproducible workflow with Structure from Motion (SfM). The applied approach for underwater SfM-documentation includes on-site workflow and post-processing.

The on-site workflow covers all steps from the preparation of the archaeological structures to the photographic data acquisition, the calculation of a preliminary 3D-model and its on-site verification. The crucial step was to ensure the suitability for modeling of the data before the situation underwater was irreversibly changed through sampling.

Post-processing was carried out in Adobe Photoshop, Agisoft PhotoScan and QGIS where the data was optimized in quality and standardized from digital image processing to the construction of a georeferenced orthomosaic. Applying these results, we can later visualize patterns in the spatial distribution of the piles concerning e.g. their age, their size or their wood species. This will lead to answers regarding architecture, internal chronology, and in-site settlement dynamics.

With this newly standardized two-step-workflow for underwater structure documentation, we are able to asses and compare the quality of each orthomosaic in a reproducible way. The presented method is highly promising for underwater-documentation of prehistoric pile-fields, yielding accurate digital plans in an efficient and cost-saving way.