A CRITERION FOR REMOTE SENSING
OBJECTS DISCRIMINATION IN CASE OF NON-GAUSSIAN DISTRIBUTION OF
SIGNATURES
A.V. Popov
The error probability is the
conventional criterion used in objects recognition while working with radar
data. It is shown that in case of non-Gaussian distribution of the objects’
signatures the error probability criterion may result in “masking” some objects
by some others. Therefore, a new criterion for signatures informativity
estimation is suggested. The given criterion is based on Kulback’s
divergence applied to the case of big amount of objects’ classes. Results of
comparative analyses of the discrimination criterion and the error probability
in case of non-Gaussian distribution of recognition-informative signatures are
presented.
Key words: remote
sensing, target recognition, error probability, Kulback’s
divergence, non-Gaussian probability distribution.