Técnicas de procesamiento
Inteligencia Artificial y algoritmos
Técnicas para el análisis y procesamiento de imágenes aplicando métodos de machine y deep learning
AI methods for image fusion
Diverse methods exploiting the statistical and geometrical characteristics of a high resolution image, to be applied to a coarser image in order to obtain an enhanced version. The methods considered include PCA extension, arithmetic algorithm, high pass filters, neuronal networks, multirresolution analysis, etc
Neural networks for data pattern detection and prediction systems
Self-Organizing Maps (SOM) is an unsupervised machine learning technique used for pattern recognition, data reduction and clustering in collections of large data sets.
SOMs is based in a Neural Network architecture. Applications include the extraction of coherent structures in fludi flows (i.e oceanic and atmospheric flows), physical and biological geographicregionalizations, phenology studies, remote sensing data gap filling, and ocean current forecast.
AI methods to estimate water masses
Unsupervised learning algorithms are used to determine the volume of water in a river dam. The method includes the determination of a mask of water-covered pixels and an estimation of the bathymetry, in order to estimate the actual volume.
Estimation of coastal bathymetry with optical imagery
Satellite visible images can be used to determine the bathymetry close to the coast and up to a few meters. Sun glint and atmospheric absorption must be corrected. Multi-scene algorithms are used to correct turbitidity. Final maps are validation with LIDAR images.
Estimation of water quality with optical imagery
Estimations of the turbidity can be used to provide water quality indices, depending on the presence of suspended matter in the water. Turbidity is assess using the spectral information of the images, relating it to the physical properties of the water column
Estimation of HAB with optical imagery
Harmful Algae Blooms (HABs) are a major concern for administrations responsible for coastal waters. Using Sentinel-2 images, new HAB indexes capable on complex waters have been developed
Object-based image analysis: precision agriculture and other disciplines
Imagery and geo-spatial data from diverse:
- Platforms (satellite, airborne, drone)
- Resolutions (spatial, spectral, temporal)
- Formats (raster, vector, LIDAR, point clouds)
are combined and /or analysed with automatic algorithms to extract information from the objects in the images
Segmentation à Classification à Export
Characterization of wetlands using LandSat imagery
Specific products based on LandSat imagery have been developed. Images are first geometrically and radiometrically corrected, then processed and classified to obtain flood indexes, turbidity, deph and hydroperiod.
