This thesis covers research into radar target characterisation and recognition with the use of joint time-frequency representations and image processing techniques. Specifically, the recognition of the scatterers of ballistic missiles is of particular interest. The complex motion and composite structure of the targets result in highly nonstationary signal, and therefore its characteristics are investigated both in time and frequency. A novel method for time-frequency analysis of nonstationary radar signals is proposed facilitating formulation of a feature space suited to the problem at hand. This signal-adaptive procedure allows unsupervised computation of the Smoothed Wigner-Ville Distribution based on time and frequency moments of the time-segmented signal.