Two-Dimensional ESPRIT with Tracking for Radar Imaging and Feature Extraction.

摘要

ESPRIT processing appears to be the best of the known spectral- analysis techniques. It provides the highest resolution and has no spectral splatter. By applying matrix eigenstructure analysis, it gives a direct answer to the direct question 'What frequencies, real or complex, are present in the data and what are their amplitudes'. Conventional Fourier techniques, as well as some of the other higher-resolution methods, answer the less direct question 'What amplitudes, applied to a set of regularly-spaced real frequencies, best represent the data'. Then comes the problem of interpreting the amplitudes. These attributes of ESPRIT, in the two-dimensional version described here, make it a natural for radar signal processing, where it answers the need for high- resolution imaging, free of sidelobes in range and Doppler, and for high- fidelity target feature extraction. For example, the uncertainty in the scatterering-center locations in an ESPRIT image extracted from high-quality static-range radar data collected over a bandwidth of 1 GHz is just a few millimeters; for conventional Fourier processing of the same data the uncertainty is many centimeters. The signature of the base edge of a perfectly conducting cone extracted from static-range data by ESPRIT agrees accurately with the signature predicted by edge-diffraction theory. This report starts with a mathematical model for the radar data, describes a technique for 'resampling' the data to achieve a more perfect fit with the ESPRIT data model, summarizes the two-dimensional ESPRIT algorithm itself, and presents examples of its performance. The appendix covers the details of this least-mean-square version of ESPRIT, including an enhancement that allows the scatterers to be tracked individually.