Hyperspectral Imaging

Core Idea

Hyperspectral sensors (imaging spectrometers) record reflected radiation in hundreds of narrow, contiguous spectral bands (typically 5-10 nm wide), producing a near-continuous reflectance spectrum for every pixel. Where multispectral sensors sample at a few strategic wavelengths, hyperspectral sensors capture the complete spectral shape, enabling identification of specific minerals, chemicals, vegetation species, and materials based on diagnostic absorption features too narrow for broadband sensors to resolve. This comes at the cost of enormous data volume and complex processing requirements.

Explainer

Multispectral imaging samples the spectrum at a handful of strategic points. Hyperspectral imaging measures a near-continuous spectrum for every pixel, typically in 100-300 bands each 5-10 nm wide, spanning 0.4-2.5 um. The result is an image cube -- two spatial dimensions plus one spectral dimension -- where each pixel contains a complete reflectance spectrum.

The scientific motivation is material identification through diagnostic spectral features. Many minerals, chemicals, and biological materials have absorption features narrower than 50 nm that multispectral sensors cannot resolve. Iron oxides, carbonates, sulfates, and clay minerals each have characteristic SWIR absorptions. Vegetation species differ in subtle features related to leaf chemistry. Water quality parameters each affect the spectrum in distinct narrow-band ways.

The processing pipeline is substantially more demanding. Atmospheric correction must be accurate for each narrow band. Dimensionality reduction techniques (PCA, minimum noise fraction) compress hundreds of correlated bands into meaningful components. Spectral matching compares each pixel against known material spectra. Spectral unmixing estimates fractional abundance of multiple materials within a single pixel.

Current missions include PRISMA, DESIS, and EnMAP, with NASA's SBG planned for global coverage. The technology is also widely deployed on aircraft for targeted campaigns. The trend is toward making hyperspectral data increasingly accessible, but multispectral imaging remains the backbone of operational remote sensing due to simpler processing and longer archives.