THE USE OF LINEAR AND NON-LINEAR MAPPING METHODS TO RESOLVE PROBLEMS IN THE CLASSIFICATION OF REMOTE SENSING DATA
The paper reviews recent developments in the use of LINEAR MAPPING of original remotely sensed data into more effective features, from the point of view of class separability. Different theoretical investigations have been described and compared. This paper also illustrates the importance, in multihypothesis Pattern Recognition problems, of the connection between a NON-LINEAR MAPPING into the space defined by the a-posteriori probability distributions (corresponding to the classes under investigation) and the Bayesian Probability of misclassification.
Bibliographic Reference: EUR 9131 IT (1984) MF, 37 P., BFR 120, BLOW-UP COPY BFR 185, EUROFFICE, LUXEMBOURG, POB 1003
Availability: Can be ordered online
Record Number: 1989122103200 / Last updated on: 1987-01-01
Available languages: it