Unit Circle Roots Constrained Subspace-Based Adaptive Matched Filter Detector for Moving Target Detection with Limited Secondary Data

Since the estimation accuracy of the sample covariance matrix (SCM) is limited by the secondary data number, the filter of the subspace-based adaptive matched filter (SAMF) detector constructed by the SCM has insufficient null depth for clutter and noise, which affects the detection performance of moving targets. To address this issue, a unit circle roots constrained SAMF (UCRC-SAMF) detector is proposed. This article demonstrates that, with a clairvoyant covariance matrix, the roots of the z-polynomial of the SAMF’s filter are located on the unit circle. The proposed method leverages this property to enhance the filter constructed by the SCM. First, a linear transform is applied to the filter by cascaded Givens rotation matrices. Second, the z-polynomial of the transformed filter is factorized, and the root of each factor is moved onto the unit circle by conjugate symmetry processing. This step compensates for the inaccuracies introduced by the limited secondary data. Third, this method iteratively adjusts the Givens rotation matrices in the first step to ensure that the filter satisfies the scale constraint of the detector. Comparisons based on both simulation and real data show that the constrained filter achieves a deeper null region for the clutter and noise, and the proposed UCRC-SAMF detector can improve the detection performance with limited secondary data.