Automating Archaeological Feature Detection: Unsupervised Classification and Feature Extraction from Satellite Imagery

Summary

This is an abstract from the "SAA 2023: Individual Abstracts" session, at the 88th annual meeting of the Society for American Archaeology.

Satellite and aerial images are used for archaeological site prospection worldwide. However, manually detecting and mapping archaeological sites from imagery can be time consuming. This poster examines the utility of an image processing and unsupervised classification procedure for archaeological feature detection and mapping in arid settings. This procedure, built in Python, utilizes several geospatial analysis modules to read in, process, and classify raster images of archaeological sites. Features are detected and vectorized, and accuracy can be optionally assessed by the percentage of the total length of previously mapped features captured for a given image area. This tool is tested on Google Earth images of ancient agricultural fields, focusing on features in arid settings with little vegetation (leeward Hawaiʻi Island, U.S. Southwest). There are currently several limitations to this tool, namely false positives. However, the tool is extendable and could be applied to different archaeological feature types in different geographic areas. Although this tool requires additional refining, the accuracy at this stage suggest the potential utility of unsupervised classification for automated archaeological feature detection.

This Resource is Part of the Following Collections

• SAA 2023: Individual Abstracts   •
• Society for American Archaeology 88th Annual Meeting, Portland, OR (2023)

Keywords

Record Identifiers

Abstract Id(s): 36587.0