Project funded by the Italian Space Agency (ASI).
The high-resolution space and spectral satellites have changed the way we consider the environment and the environmental phenomena. This is the case of PRISMA (Hyperspectral Precursor of the Application Mission), a cutting-edge Earth observation system, equipped with electro-optical tools, which integrates a hyperspectral sensor with a medium-resolution camera, sensitive to all colors (panchromatic).
The PRISMA_Learn project is located as a primary call for developing advanced machine learning techniques to demonstrate the capabilities of the brand new PRISMA mission and its relevance to address the modern challenges associated to the Earth observation. The project is therefore finalized to development of innovative machine learning algorithms and experimental validation through PRISMA data and their integration with other missions.
During the project four main strategic problems will be addressed:
- Mapping of subtle land cover classes (e.g., different materials in urban areas, vegetation species in cropped areas) by fusion of hyperspectral and panchromatic data.
- Unmixing of hyperspectral data with explicit characterization of non linear interactions between the specific endmembers spectral signatures in urban scenarios.
- Change detection in bi-temporal and multi-temporal sequences of images, with focus on detecting subtle changes in cropped areas.
Change detection from heterogeneous data sources, especially mixing PRISMA and COSMO-SkyMed 2nd generation images.
