High resolution 3D "snapshot" ISAR imaging and feature extraction

We develop a new formulation here for three dimensional (3D) radar imaging of inverse synthetic aperture radar (ISAR) data based on recent developments in high resolution spectral estimation theory. Typically for non real-time applications, image formation is a two step process consisting of motion determination and image generation. The technique presented focuses on this latter process, and assumes the motion of the target is known. The new technique offers several advantages over conventional techniques which are based on the correlation imaging function. In particular, the technique provides for a direct 3D estimate (versus back projection to a 3D target grid matrix) of the locations of the dominant scattering centers using only a minimum set of independent 2D range-Doppler ISAR "snapshots" of the target. Because of the snapshot nature of the technique, it is particularly applicable to 3D imaging of sectors of sparse-angle data, for which the sidelobes of the correlation imaging integral become high. Furthermore, the technique provides for an estimate of amplitude and phase of each scattering center as a function of aspect angle to the target, for those aspect angles which encompass the set of 2D range-Doppler snapshots. Thus, for example, for a spinning target, the characteristics of each resolved scatterer present on the target as a function of roll angle are readily identified. The technique gracefully reduces to a range-only (e.g., wideband radar, low pulse repetition frequency (PRF)) and Doppler only (e.g., narrowband radar, high PRF) 3D imaging capability. By implementing a range-Doppler correlation tracker on a sequence of pulses, one can isolate the 3D motion of any specific scatterer using the simple 2D to 3D mapping developed in the paper, and extract the scatterer amplitude and phase as a function of time. Using this scatterer response, combined with the 3D scatter locations extracted from the 3D image, one can directly reconstruct the actual scattering measurements using a generalized point scatter model based on the Geometrical Theory of Diffraction (GTD). Correlating the extracted GTD-based diffraction coefficients to a library of hypothesized characteristics of the scatterer response provides the potential of enhanced scatterer typing and identification. Results illustrating the technique developed are presented for both simulated and static range data.