Target Classification Based on a Combination of Possibility and Probability likelihood in the Bayesian Framework

Abstract

The problem of target classification is addressed in the Bayesian framework as an interpretation of the likelihood of Bayes’ theorem as a possibility. A better explanation and definition based on this perspective, as opposed to the conventional probability interpretation, is given for the uncertain mapping from the class space to the feature space. In this manuscript, we propose a new Bayesian classifier that can naturally combine both the probability and the possibility through a reinterpretation of Bayes’ theorem. An example of target classification using kinematic features demonstrates that the proposed Bayesian classifier outperforms the conventional Bayesian classifier and gives accurate classification results.