Spectral Angle Mapper

Spectral Angle Mapper evaluates similarity geometrically: each measured spectrum and reference spectrum becomes a vector, and their angular separation is calculated. A smaller angle indicates stronger spectral agreement, so the method can distinguish candidate materials according to how closely their signatures align rather than relying on signal magnitude alone. This provides a direct basis for supervised classification in hyperspectral data.

The angular comparison reduces sensitivity to overall signal intensity. Consequently, spectra with similar patterns can remain comparable even when their measured signal levels differ. This feature is important when classification should emphasize the shape or directional relationship of spectral signatures instead of treating a uniformly stronger or weaker signal as evidence of a different chemical constituent or material.

Reference spectra provide the comparison standards for identifying unknown measurements, so they need to represent the materials or chemical constituents being sought. Threshold settings then determine how much angular difference is acceptable for classification. Poorly representative references or unsuitable thresholds can reduce the usefulness of the resulting classification, even when the angular calculations themselves are performed correctly.

A typical workflow begins with measured hyperspectral signatures and a set of reference spectra. The technique treats each spectrum as a vector, calculates the angle between each unknown and the relevant references, and uses the resulting similarity values with selected thresholds. The classified results can then be interpreted as spatially distributed material or constituent patterns.

SAM can support compositional mapping, quality assessment, and analysis of samples distributed across space. By comparing measured signatures with references, researchers can examine where particular chemical constituents, minerals, or other materials occur and evaluate sample-related patterns. Its value therefore extends from simple identification to interpreting composition and spatial variation within hyperspectral images.

In chemistry, hyperspectral measurements may contain signatures associated with multiple constituents across a sample or image. SAM offers a way to compare those measurements with reference spectra and organize the results into material or constituent classifications. This supports chemical interpretation when researchers need to connect spectral information with composition, quality, or the spatial distribution of analyzed samples.