Improved detection of close proximity targets using two-step NHD

Abstract

In airborne radar adaptive signal processing, the covariance matrix is usually estimated using secondary (training) data cells taken from adjacent range cells located symmetrically around the test cell. In non-homogeneous clutter, many of these data cells may lack the IID property, resulting in estimation performance loss. Nonhomogeneity detectors have been introduced in order to achieve more representative data selection. The generalized inner product (GIP) has been shown to work well with measured data. In this paper, we introduce a variation of the GIP to filter out the non-representative data. Moreover, the proposed approach makes use of equalized data based on the GIP. Results using the MCARM database show improved performance.