Fusion methods for multisensor classification of airborne targets

This paper tackles the problem of multisensor classification in a ground-air defence context. In order to solve it, we propose firstly to list the various features which may be extracted from the sensors generally used in such a configuration and which can be exploited for the classification. Next we deal with two categories of methods, namely: the Bayesian method and fuzzy classification methods. The latter have been selected since they belong among the most widely employed data association techniques. Additionally, the Bayesian method has the attraction of exhibiting a well-known formalism but its main drawback is that it takes no account of the imprecision which is sometimes encountered in the knowledge of the classes. By contrast, fuzzy methods make it possible, on the one hand, to associate uncertain and imprecise quantities and, on the other hand, to treat particular cases of classification in which no learning base is available. Several simulations relating to the above problem are presented. They demonstrate the advantages and drawbacks of the various methods proposed and enable their respective performance to be studied. A further avenue of research could consist in validating these methods on the basis of actual measurements recorded with the aid of several different sensors.