Day 4

Detailed paper information

Back to list

Paper title Detection of archaeological sites hidden by a forest canopy using Sentinel-1 time-series
Authors
  1. Florent Michenot SONDRA, CentraleSupélec Speaker
  2. Régis Guinvarc’h SONDRA, CentraleSupélec
  3. Laetitia Thirion-Lefevre SONDRA, CentraleSupélec
Form of presentation Poster
Topics
  • D2. Sustainable Development
    • D2.12 Cultural and Natural Heritage
Abstract text Sentinel-1, the SAR satellite family of the Copernicus program, provides the scientific community with global and recurring Earth Observation data for free. However, SAR images are subject to speckle, a form of noise that makes visual interpretation difficult.
By compensating for this drawback and leveraging the strengths of SAR imaging, it is possible to detect structures hidden by a forest canopy, even when optical imagery yields no results.

Speckle is generally reduced using spatial techniques, like multi-looking or spatial filtering. However, they decrease the (already poor) spatial resolution of the picture. Temporal speckle filtering is an alternative. A temporal mean over a (small) stack of images of the same scene will drastically reduce speckle without any degradation of the spatial resolution. Large enough structures should then be visible even when under a forest canopy.
Additionally, when trying to detect buildings, the contrast between even small static structures and variable targets (like the forest canopy) is increased. This further demonstrates that in this context, temporal speckle filtering is an improvement on spatial filtering.

By then computing the difference between the ascending and descending points of view of Sentinel-1, it is possible to further highlight hidden buildings. The technique will color western and eastern facing parts of structures and terrain (i.e., positive and negative differences) using a different color. Flat horizontal surfaces (i.e., near-zero difference) will also appear with a different color.

The technique was used over several known archaeological sites in the Guatemalan jungle. Nakbe, in particular, illustrates well the value of our method. Optical images show little indication of the presence of structures, except possibly the top of one of the pyramids and what might be a clearing. Once processed, the SAR image reveals quite clearly two large buildings and several small ones.
A map of the site found in an archaeological paper confirms the presence and positions of the structures, but also that not all are detected. This may be due to the state of conservation of the different buildings: the map might be representing the site as it was when it was built instead of as it is now.