Radar signature prediction and feature extraction using advanced signal processing techniques

Novel radar signature prediction and feature extraction techniques are proposed for applications in automatic radar target recognition. In radar signature prediction, interpolation and extrapolation algorithms are developed for fast electromagnetic solvers to reduce the computation time incurred by multiple frequency or aspect computations. The essential idea is to parameterize the induced current on the target surface using a multiple-arrival model. Based on the assumption that all of the unknown samples to be interpolated or extrapolated obey the same reduced-order model as the known samples, the model coefficients can be estimated using signal processing techniques. For parameter estimation, the super-resolution algorithm ESPRIT is used in the extrapolation process and the adaptive feature extraction algorithm is used for interpolation. After all of the model parameters are determined, the induced current and scattered fields are interpolated or extrapolated in frequency and angular domain to give values at the unknown sample indices. Various numerical examples are studied. The algorithms are also applied to efficiently predict the inverse synthetic aperture radar (ISAR) image of the benchmark VFY218 airplane in the UHF band. The results agree well with anechoic chamber measurement data.; For signature feature extraction from measured radar data, a motion compensation algorithm and a Doppler artifact extraction algorithm are developed. The adaptive joint time-frequency technique is applied to track the time history of the Doppler frequency components due to the motion of the target or additional moving components on the target. The technique entails an iterative search-and-extract procedure in the joint time-frequency plane to find the chirp bases that best represent the time-frequency behavior of the signal. According to the parameters of the chirp bases, motion effects are determined or classified. For moving targets, the motion parameters of the target are extracted and the bluffed ISAR image due to motion errors can be refocused. For targets with fast rotating components, the Doppler artifacts induced by the rotating parts are separated from the ISAR image of the target body. In addition, the motion features of the rotating parts can be extracted and interpreted. The algorithms are verified on simulated radar data. They are also successfully applied to measured radar data of airplanes and helicopters in flight.