Detection and characterization of landslides
Organizations involved: UBO (France)
The workflows targets to detect slow-moving landslides from a time-series of Very High Resolution satellite images. Slow-moving landslides are widespread in many landscapes with significant impacts on the topographic relief, sediment transfer and human settlements, whereas their area-wide mapping and monitoring in mountainous terrain is still challenging. The growing archives of optical remote sensing images offer a great potential for the operational detection and monitoring of surface motion in such areas. This workflow provides tools for measuring piece‐wise locally the shift among two or more co-registered datasets provided in GeoTIFF format.
The main steps of its processing chain consist in the tiling of Geospatial raster data, the measurement of 2D displacement measurement, the Merging of Geospatial raster data and a spatio-temporal filtering of displacement time series.
The result of the workflow is three thematic map (raster) showing X and Y displacement fields and a correlation coefficient for a more accurate detection and quantification of dunes displacement.
To assess the performance and guide further development, this workflow has been compared against state‐of‐the art image matching methods in terms of quality of the results, Figure 8, scalability of the computation and the required amount of user interaction.
The IQmulus algorithms have been tested on a landslide study around “Mare à Poule d’Eau” (Hell-Bourg, Reunion island) conducted by Réjane Le Bivic, UBO (thesis in progress). The Spot 5 satellite data covers approximately 15 sq.km, (2*1GB).The processing on the Fraunhofer cloud using took around 7 hours. The accuracy of the landslide detection was around 99% compared to Le Bivic et al. methodology (thesis, in progress). On the figure bellow the landslide results with accuracy assessment is represented.
