Glossary for Geospatial Science

maximum-likelihood classification:� A supervised image processing routine usually applied to multivariable images that have a dimensionality greater than 3.� First, the user selects sample areas, (called prototype or training set areas) for each feature to be mapped.� The maximum-likelihood process then computes the statistical properties of these features.� Similarly, it determines the statistical characteristics which separate feature types.� After this identification model has been built, the process tests each cell to determine in which prototype group it most probably belongs.� A threshold can be set so that if the highest probability of a feature match is below the threshold, the cell is designated as unknown.� This �catch-all� group insures that a cell representing a feature type not defined in one of the training sets will not be forced into membership.