Automated Qanat Detection: Examining the Application of Deep-Learning in Archaeological Remote Sensing

Summary

This paper presents the preliminary results of a collaborative project that seeks to develop a deep learning model for automated detection of qanat shafts on CORONA Satellite Imagery. Increasing quantity of air and space-borne imagery available to archaeologists and advances in computational science has created an emerging interest in automated archaeological detection. Previous studies have applied machine learning algorithms for detection of archaeological sites and off-site features, with varying success rates. In the last few years, tremendous success has been achieved in image recognition through deep learning, a category of supervised machine learning which is based on hierarchical representation learning. We have chosen to examine the application of deep convolutional neural networks for automated archaeological detection focusing on qanats. The methodological merit of the project is that qanat shafts are one of the most suitable archaeological features for pattern recognition. The analytical merit is that conducting comparative studies of qanat systems at a regional scale is predicated on mapping massive number of qanat shafts, which is impossible manually. Processing of big data generated through machine learning would allow us to examine how this sustainable water supply technology has been adapted to a remarkably wide range of water deficit environments.

This Resource is Part of the Following Collections

• Society for American Archaeology 83rd Annual Meeting, Washington, DC (2018)

Keywords

Spatial Coverage

Record Identifiers

Abstract Id(s): 20402