Indicative applications include, but are not limited to:
Delineation of height categories for vegetated areas characterization through texture analysis of a single very high spatial resolution multispectral image.
A number of texture features are produced, including local variance, entropy, and binary patterns. These features are processed through a variety of machine learning algorithms, potentially including dimensionality reduction, feature selection,
multiple imputation of missing data, outlier removal, data normalization. Following processing, each land patch is assigned to the respective height category through a number of different supervised classifiers.
Characterization of habitats based on land cover properties of patches and very high resolution satellite imagery. Having the land cover class of each patch expressed in the Land Cover Classification System (LCCS)
as input, a number of spectral, texture, topological, morphological, and height features are extracted. The features are used to assign each patch to a habitat class expressed in the General Habitat Categories taxonomy,
following either a supervised classification or a fuzzy evidential reasoning rule-based scheme taking into account noise afflicted data and uncertainty.
Extraction of biodiversity indicators through the use of remote sensing data. A number of indicators adopted by the
European Union and international organizations are estimated through the use of remote sensing data in an accurate and timely manner.
Mapping and assessment of land cover; Remote Sensing data constitute precious information for the understanding of land use and land cover (LULC) of various territories.
eos team is actively engaged in LULC mapping activities, change detection, precision agriculture, monitoring of land surface processes and validation. Recent tangible results of team's work include:
- Hydroperiod Estimation (HydroMap). Generates HydroMaps from series of water masks, falling within the desired time-period.
- Inland free water surface derivation from Sentinel-2 satellite imagery (WaterMasks). Generates water masks following the unsupervised local thresholding approach.
- Landscape Fragmentation measures calculation (LandMetrics). Calculates a number of landscape measures used as indicators of fragmentation and/or connectivity of land cover or habitat classes in the selected study area.
- Sentinel-1 data speckle noise suppression (SpeckleRemoval). Suppresses speckle in the SAR Sentinel-1 product (developed for GRD data) by using guided image filtering.
- Detection of changes in NDVI approximated phenological cycles (PhenologyChanges). Estimates the abrupt changes that occur during a range of years, based on a provided time-series of NDVI (Normalized Difference Vegetation Index) GeoTiff files. BFAST CRAN package is used.
- Estimation of phenology metrics (PhenologyMetrics). Calculates the Greenup, Senescence and Max NDVI (Normalized Difference Vegetation Index) day of a collection of NDVI GeoTiff files within a season, based on phenex R package. Application can detect multiple phenological cycles.
You may find more information for each module here (click on the tab "Workflows" and then select the module that you wish).