Machine Learning: a Toolkit for Speeding Up Archaeological Stratigraphic Identification

Digging a site, recording the stratigraphic units and interpreting the results in order to comprehend the historical processes of the site formation are part of archaeological excavation work. As archaeologists dig, they consider the extension, color, texture, hardness, and composition of the soil that they are removing. These processes are timeconsuming, and may be affected by human skill. The main idea of this work is to automatize stratigraphic unit detection and characterization. To this end, a Machine Learning algorithm has been applied to digital images of archaeologic excavation sites for classifying regions that are similar in color and the contours of which represent stratigraphic units. Each stratigraphic unit has been characterized in terms of texture according to the mean energy. This combined approach speeds up the documentation work: since the results are readily digitalized during an excavation, they could offer a prompt guide for archaeologists.