A Doppler-based target classifier using linear discriminants and principal components

This paper describes the design of the automatic target classifier which has been introduced into the AMSTAR battlefield surveillance radar. It discusses the requirements which have driven the design of the classifier, the data which is used to make the classification, the choice of linear discriminant analysis as the classification technique and the use of principal components analysis to simplify the training of the discriminator. It also discusses the testing of the classifier and the performance achieved.