Analysis of clutter cancellation in the presence of measured L-band radar ground clutter data

A well-known radar signal processing problem is the detection of a target signal having known form in the presence of correlated clutter and thermal noise. When the signal to be detected is embedded in correlated Gaussian distributed clutter, the optimum Neyman-Pearson detector is the linear whitening matched filter (MF). The contribution of the present paper is to investigate, by means of L-band measured ground clutter data, the robustness of the linear matched filter operating in a Gaussian environment in the presence of a mismatch between the design clutter power spectral density (PSD) shape and the actual shape. The well-known Gaussian and power-law PSD are compared to the exponential PSD that has been revealed by experimental measurements carried out by the MIT Lincoln Laboratory (MIT-LL) Phase One and LCE (L-Band Clutter Experiment) coherent radars on ground clutter data. The parameters of these three models are estimated by means of a nonlinear least squares (NLLS) method. The impact of the spectral models on the performance of the matched filter is investigated in terms of improvement factor (IF), probability of false alarm and probability of detection. The numerical results of this paper validate the exponential clutter spectral model for windblown foliage by showing that the differences between using actual measured in-phase and quadrature clutter data and modeled clutter spectral data of various spectral shapes are minimized when the spectral model employed is of exponential shape. Our conclusions are summarized.