Preprint Article Version 1 NOT YET PEER-REVIEWED

Assessment of Hyperspectral Indicators for the Detection of Buried Archaeological Remains

  1. German Aerospace Center (DLR), Remote Sensing Technology Institute (IMF), 82234 Wessling, Germany
  2. Cyprus University of Technology, Department of Civil Engineering and Geomatics, Saripolou 2-8, 3036 Limassol, Cyprus
  3. Laboratory of Geophysical-Satellite Remote Sensing and Archaeo-environment, Institute for Mediterranean Studies, Foundation for Research and Technology, Hellas (F.O.R.T.H.), Crete, Greece
Version 1 : Received: 21 October 2016 / Approved: 22 October 2016 / Online: 22 October 2016 (10:37:39 CEST)

How to cite: Cerra, D.; Agapiou, A.; Sarris, A.; Hadjimitsis, D. Assessment of Hyperspectral Indicators for the Detection of Buried Archaeological Remains. Preprints 2016, 2016100093 (doi: 10.20944/preprints201610.0093.v1). Cerra, D.; Agapiou, A.; Sarris, A.; Hadjimitsis, D. Assessment of Hyperspectral Indicators for the Detection of Buried Archaeological Remains. Preprints 2016, 2016100093 (doi: 10.20944/preprints201610.0093.v1).

Abstract

Hyperspectral images can provide valuable information on the potential presence of buried archaeological structures through the detection of crop marks. Whenever crop marks are not directly visible to the human eye, buried archaeological remains outlines can be spotted in hyperspectral data usually indirectly, by comparing the vegetation health status with its immediate surroundings. In the last years, several indicators have been used in the literature, with varying success, to improve the interpretation of satellite images for archaeological research. These might range from isolated image element values in single spectral bands to band ratios and more complex spectral indices, usually related to vegetation health. However it is difficult to assess these indices quantitatively and select the most meaningful for a given task, in particular for a non-specialist in image processing. This paper presents a first effort at ranking objectively the suitability of hyperspectral features used for the detection of buried archaeological structures in agricultural areas based on information theoretical measures. The ranking is achieved by computing the statistical dependence between the extracted features and a digital map indicating the presence of buried structures using information theoretical notions. The reduction of uncertainty in the former by observing the latter is quantified through mutual information, which excels at measuring non-linear correspondences. Two case studies are reported: the Roman buried remains of ancient Carnuntum (Austria), and the Nelotithic tells in the Thessalian plain (Greece), where respectively airborne and spaceborne hyperspectral acquisitions are available. First results indicate that such approach may aid archaeologists by steering the visual analysis of multidimensional data selecting the best band combinations that maximize the separation between structures of interest and background.

Subject Areas

hyperspectral imaging; archaeology; cultural heritage; mutual information; spectral indices; crop marks

Readers' Comments and Ratings (0)

Discuss and rate this article
Views 136
Downloads 84
Comments 0
Metrics 0
Discuss and rate this article

×
Alerts
Notify me about updates to this article or when a peer-reviewed version is published.