Title :
A Mahalanobis distance fuzzy classifier
Author :
Deer, P.J. ; Eklund, P.W. ; Norman, B.D.
Author_Institution :
Div. of Inf. Technol., Defence Sci. & Technol. Organ., Salisbury, SA, Australia
Abstract :
Traditional `hard´ classification techniques are inappropriate for classifying remotely sensed imagery. Class `boundaries´ in the natural environment are not distinct and a single pixel may exhibit spectral characteristics related to a number of classes. Fuzzy set theory was introduced to address the issue of the `vagueness´ of class or set membership. An unsupervised approach to fuzzy classification uses the fuzzy c-means algorithm. The paper reports on a related supervised approach in which training sets are selected, then the fuzzy class memberships are determined by the reciprocal of the Mahalanobis distance from these training class means
Keywords :
fuzzy set theory; image classification; learning (artificial intelligence); remote sensing; Mahalanobis distance fuzzy classifier; class boundaries; class vagueness; fuzzy c-means algorithm; fuzzy class memberships; fuzzy set theory; natural environment; pixel; remotely sensed imagery classification; set membership vagueness; spectral characteristics; supervised fuzzy classification; training class means; training set selection; Classification algorithms; Clouds; Computer science; Fuzzy set theory; Fuzzy sets; Image classification; Image processing; Information technology; Pixel; Remote sensing;
Conference_Titel :
Intelligent Information Systems, 1996., Australian and New Zealand Conference on
Conference_Location :
Adelaide, SA
Print_ISBN :
0-7803-3667-4
DOI :
10.1109/ANZIIS.1996.573940
