Ship Classification by Cyclostationary Analysis of Preimage Formation ISAR Data

Abstract

Inverse synthetic aperture radar (ISAR) has proven valuable for imaging in maritime contexts. However, micro-Doppler (m-D) effects from rotating components on ships often blur ISAR imagery and complicate classification. On the other hand, m-D measurement carries unique characteristics that can distinguish specific vessel types with shared structural features. This article introduces a cyclostationary m-D framework for ship classification that fundamentally differs from prior approaches by: 1) extracting m-D features before ISAR image formation; 2) recovering physically interpretable micromotion parameters; and 3) relaxing the need for strictly sinusoidal motion models. We validate the proposed approach through a hierarchy of simulation experiments, estimating rotators’ unique locations, frequencies, and radii. The proposed approach demonstrates improved performance over traditional m-D estimation techniques such as short-time Fourier transform (STFT) in noisy environments. Our findings illustrate the robust extraction of m-D features for accurate ship classification in challenging maritime environments.