An application of advanced spectrum estimation to multi-channel radar detection and location

We explore the application of an advanced spectral estimation technique to target detection and localization in multi-channel space radar systems. Previous work had applied motion compensation and array calibration to produce SAR-like range-Doppler images for each channel, followed by a maximum likelihood angle of arrival (AOA) and alternating projections (AP) technique designed to optimally detect and locate targets in this representation. Moving targets in a given range-Doppler cell exhibit a channel-to-channel phase difference after the motion compensation and array calibration. This phase progression may also be detected and estimated using spectral estimation approaches. However, since the target and clutter phase ramps are very close, a super-resolution approach is needed. The reduced-rank linear predictor ROCKET is investigated in this role. It is shown that the original technique has a two-threshold capability that provides two opportunities to discard false alarms, whereas the pure ROCKET-based approach only has one. As a result, the two-threshold AOA+AP technique generally outperformed the ROCKET-based algorithm in Monte Carlo simulations. Furthermore, the two-threshold construct holds the promise for further performance improvements pending ongoing investigations