Automated learning multi-criteria classifiers for FLIR ship imagery classification

This paper proposes an automated learning method (ALM) based on real-coded genetic algorithm (RCGA) to infer the multi-criteria classifiers (MCC) parameters. The multi-criteria classifiers (or multi-criteria classification methods) considered are based on concordance and discordance concepts. A military database of 2545 forward looking infra-red (FLIR) images representing eight different classes of ships is therefore used to test the performance of these classifiers. The empirical results of MCC are compared with those obtained by other classifiers (e.g. Bayes and Dempster-Shafer classifiers). In this paper, we show the benefits of cross-fertilization of multi-criteria classifiers and information fusion algorithms.