Automatic detection of potential buried archaeological sites in Saruq Al-Hadid, United Arab Emirates

El Rai, Marwa Chendeb and Al-Saad, Mina and Aburaed, Nour and Al Mansoori, Saeed and Al-Ahmad, Hussain and Marshall, Stephen; Erbertseder, Thilo and Chrysoulakis, Nektarios and Zhang, Ying, eds. (2020) Automatic detection of potential buried archaeological sites in Saruq Al-Hadid, United Arab Emirates. In: Proceedings Volume 11535. Proceedings of SPIE, 11535 . SPIE, Bellingham, Washington. ISBN 9781510638846 (

[thumbnail of El-Rai-etal-SPIE2020-Automatic-detection-of-potential-buried-archaeological-sites]
Text. Filename: El_Rai_etal_SPIE2020_Automatic_detection_of_potential_buried_archaeological_sites.pdf
Accepted Author Manuscript

Download (1MB)| Preview


The use of remote sensing in archaeological research allows the prospection of sub-surfaces in arid regions non- intrusively before the on-site investigation and excavation. While the actual detection method of expected buried archaeological structures is based on visual interpretation, this work provides a supporting archaeological guidance using remote sensing. The aim is to detect potential archaeological remains underneath the sand. This paper focuses on Saruq Al-Hadid surroundings, which is an archaeologist site discovered in 2002, located about 50 km south-east of Dubai, as archaeologists believe that other archaeological sites are potentially buried in the surroundings. The input data is derived from a combination of wavelength L-band Synthetic Aperture Radar (ALOS PALSAR), which is able to penetrate the sand, and multispectral optical images (Landsat 7). This paper develops a new strategy to help in the detection of suspected buried structures. The data fusion of surface roughness and spectral indices enables tackling the well-known limitation of SAR images and offers a set of pixels having an archaeological signature different from the manmade structures. The potential buried sites are then classified by performing a pixel-level unsupervised classification algorithm such as K-means cluster analysis. To test the performance of the proposed method, the results are compared with those obtained by visual interpretation.