Spectral angle classification

In spectral angle classification, a set of reference spectra represent the classes into which the image is to be partitioned. In this process, N-band spectra are interpreted as N-band vectors. At each image location the normalized inner product of each reference spectra and the image spectrum at that location is computed (the normalized inner product is the cosine of the angle between the two operand vectors). The class of the reference spectrum that makes the smallest angle with the image spectrum is assigned to that location. If the smallest angle is greater than a user-specified threshold, then no class is assigned to that location.

SAM performs a spectral angle classification on an image.

Spectral angle classification is most useful when the image is mostly comprised of pixels that correspond to terrain elements consisting of a single, spectrally distinct material class of interest (that is, when the pixels are "pure").

Spectral angle classification is used when the reference spectra and the image spectra represent the same physical quantities and units. Typically, this means that the image data have been radiometrically corrected to reflectance, and that the reference spectra are ground or laboratory measured reflectance spectra.