Classification Methods for Remotely Sensed Data

by Paul Mather, Brandt Tso

The extraction of thematic information from remotely sensed images is a key area of research into applications of remote sensing data. Standard methods of classification based on similarity ond probability measures, such as the maximum likelihood procedure, are now being superseded neural/connectionist and artificial intelligence algorithms. Concepts such as fuzzy decision rules and soft classification are extending the traditional boundaries of pattern recognition. This book provides a survey of these new methods, together with a guide to essential preliminaries such as sampling methods, orthogonal transforms, feature selection, and accuracy assessment. Methods of quantifying texture using fractals and other techniques, of estimating context using Markov random fields and incorporating these features in multi-source classifications, are described in the later part of this book.

Kindle Edition

First published November 30, 2001