Geometric Method based Covariance Matrix Estimator with Application to Radar Target Detection

Abstract

Accurate covariance matrix estimation is important to radar target constant false alarm rate (CFAR) detection. In this paper, a 2-step method derived from geometric framework independent of clutter statistical distribution is employed to estimate the clutter covariance matrix. Firstly, the covariance matrix with Toeplitz Hermitian positive definite (THPD) structure of each reference cell is calculated by the correlation coefficient of the received data. Then the geometric barycentre of THPD matrix is estimated using different geometric measures on the matrix manifold. Furthermore, we obtain several modified adaptive normalized matched filter (ANMF) methods, which combine the proposed matrix estimation with the classical ANMF method. Finally, the effectiveness of the improved ANMF methods is verified through simulation experiments, which also indicate the reliability of the proposed clutter covariance matrix estimation.