Radar target identification using one-dimensional scattering centres

Abstract

Identification concerning different types of radar targets can be achieved by using various radar signatures, such as one-dimensional (1-D) range profiles, 2-D radar images, and 1-D or 2-D scattering centres on a target. To solve the target identification problem, the authors utilise 1-D scattering centres, which correspond to the highest peaks in the 1-D range profile, The proposed approach obtains scale and translational-invariant features based on the central moments from the distribution of the 1-D scattering centres ori the target; these 1-D scattering centres can be extracted from various techniques such as inverse fast Fourier transform (IFFT), fast root-multiple signal classification (fast root-MUSIC), total least squares-Prony (TLS-Prony), generalised eigenvalues utilising signal subspace eigenvectors (GEESE), and the matrix-pencil (MP) algorithm. The information redundancy contained in these features, as well as their dimensions, are further reduced via the Karhunen-Loeve transform, followed by adequate scaling of the computed central moments. The resulting small dimensional and redundancy-free feature vectors are classified using the Bayes classifier. Finally, this new strategy for radar-target identification is demonstrated with data measured in the compact range facility, and the above five different techniques for 1-D scattering centre extraction are compared and investigated in the context of tat-get identification.