An integrated approach to passive target classification

A comprehensive study of the issues involved in passive target classification is made. A quantitative procedure based on information theoretic concept is developed to evaluate the efficacy of the features to help in making correct decisions. The three well-known classification schemes, viz., the statistical classification, the expert system and the neural network approaches are compared. The mutual information which gives a measure of the probability of misclassification is chosen as the main criterion for evaluating and comparing the performances. Finally, by combining the three schemes, an integrated passive classifier is proposed