This thesis is devoted to the analysis of flying target signatures at low frequencies, by means of a HF-VHF multifrequency and multipolarisation radar called MOSAR. Target recognition capabilities have been assessed by studying target radar cross-section (RCS) behaviour in the HF-VHF frequency band where targets cannot be steathly.Firstly, the MOSAR radar abilities are described as well as the improvements and changes introduced for an optimal operation. Then, measurement campaigns are analysed and emphasise is put on the difficulty to associate a signature to a given target without knowledge of it flight route.Secondly, the aircraft modelling possibilities are studied to better characterise RCS of targets at the operating frequencies. A commercial aircraft model has been designed and evaluated by comparison with measurements made in an anechoïc chamber. This allows the study of the influence of the flight route on aircraft RCS signature, and the results confirm the observations.Finally, target classification possibilities are considered. Two classification techniques are investigated : the nearest neighbour and the multilayer perceptron neural network (MLP). It is shown that the set of features and the number of frequencies used by the radar must be correctly chosen to minimise the probability of misclassification. Likewise, the use of these methods with MOSAR experimental data is discussed.
