Automatic Identification of Shipwrecks Using Digital Elevation Data and Deep Learning

Summary

This is an abstract from the "SAA 2021: General Sessions" session, at the 86th annual meeting of the Society for American Archaeology.

The objective of this project was to create a deep learning model that uses digital elevation data to automatically identify shipwrecks. The model uses a convolutional neural network architecture and has a F1 score of 0.92. Deep learning modeling based on remotely sensed imagery is a rapidly expanding area of research within the field of computer science, but deep learning is far less common in archaeology. Applications of deep learning are even more rare in the field of underwater archaeology. The shipwreck model, which is being completed under the aegis of the Navy’s Underwater Archaeology Branch (UA), is based on open source topo-bathymetric data and shipwreck data available from NOAA’s Data Access Viewer and NOAA’s Wrecks and Obstructions Database, as well as UA’s records. This model may help the Navy’s Office of Underwater Archaeology (UA) to find new features, create more accurate and complete maps of shipwreck locations, study patterns across the landscape, and aid management objectives. This model can easily be adjusted to identify other types of features, even using multispectral or RGB imagery as input. This work seeks to make machine learning methods accessible to non-computer scientists interested in study, management, and conservation of the landscape.

This Resource is Part of the Following Collections

• Society for American Archaeology 86th Annual Meeting, Online (2021)   •
• SAA 2021: General Sessions

Keywords

Spatial Coverage

Record Identifiers

Abstract Id(s): 33284