Unsupervised classification
How does unsupervised classification work?
Unsupervised classification, uses multi-variate clustering algorithms to group with similar spectral reflectance signatures.
Field data or knowledge are then used to assign a mapping category label to each cluster.

Unsupervised classification sequence: A) Corrected image, B) Unsupervised classification based on pixel statistics, C) Field data for assigning mapping categories to clusters of pixels or for validation, and D) Labelled unsupervised classification
References:
Jensen, J. R. (2005). Introductory digital image processing: a remote sensing perspective. Third edition, Prentice Hall: 316.