Constant False Alarm Rate Reduced Rank Detection Using the Multistage Wiener Filter

In airborne radar, space-time adaptive processing is computationally demanding and requires training data to estimate the clutter-plus-noise covariance matrix. However, in many practical environments, the available training data is often insufficient. Subsequently, reduced-rank adaptive processing techniques have been employed to overcome this problem and offer lower computational costs. The multistage Wiener filter is an efficient reduced rank adaptive detector that achieves excellent compression of the clutter subspace. However, the challenge of implementing this reduced rank detector lies in setting the detection threshold to ensure the test statistic maintains a constant false alarm rate (CFAR). In this paper, we explore the CFAR characteristics of the multistage Wiener filter detector and propose several methods to restore the detector's CFAR behaviour. The proposed methods allow the detector's threshold to be set analytically at very low computational cost.