One of the current challenges in Earth Observation is the successful application of ML approaches for the extraction of relevant information from a large variety of sensors. The new Italian PRISMA mission (PRecursore IperSpettrale della Missione Applicativa) will provide a new free dataset for research purposes. The application of the latest developments in AI provides the potential to exploit hyperspectral data and their unmatched spectral resolution more efficiently.
This research activity is the result of a collaboration between ASI and ESA. The collaboration started in November 2019 and will mainly focus on the exploitation of PRISMA data, although it is a potentially useful application in other new missions such as Φsat-1.
The aim of the research activity is to exploit hyperspectral data by means of artificial intelligence techniques. This can be carried out in two different ways:
Hyperspectral images from PRISMA enabled the estimation of very important environmental variables (vegetation stress, water resources) using various post-processing methods. In particular, hyperspectral imaging demonstrated better discrimination than multispectral instruments, thanks to its fine and continuous spectral information. Both hyper- and multi-spectral sensors record radiance in the Visible to Near-Infrared (VNIR) and Short-Wave Infrared (SWIR) of the spectrum, VNIR spanning 400–1000 nm and SWIR 1000–2400 nm. Unlike multispectral sensors, which record in a fairly limited number of discrete spectral bands (4–20 bands), hyperspectral sensors include a very large number of contiguous and narrow spectral bands of 5–15 nm.
