A Target-Focused Reduced-Rank Detection Approach

In airborne radar, target detection performance can be degraded by limited sample support in a non-homogeneous and non-stationary environment. The multistage Wiener filter (MWF) is a reduced rank adaptive detector that compresses the clutter-plus-noise (interference) subspace. However, there is a challenge in selecting the appropriate number of MWF stages to achieve these benefits. Current approaches estimate the rank of the full clutter subspace. However, this is known to be an overestimate of the required number of MWF stages that leads to a degradation in the detection performance. Consequently, we reformulate the problem and propose an algorithm that seeks the minimum number of stages required to achieve high probability of detection. The proposed algorithm utilises the Kullback-Leibler divergence to estimate the number of stages. Our proposed method leads to early truncation of the MWF, reducing sample support requirements and decreasing computational cost. We provide a computationally efficient implementation embedded within the MWF and demonstrate its effectiveness in a variety of space-time adaptive processing simulations.