Classification of UAV-to-ground vehicles based on micro-Doppler signatures using singular value decomposition and reconstruction

Abstract

Classification of ground vehicles from the unmanned aerial vehicles (UAVs) is the basis for local precision strikes. In this paper, a method based on micro-Doppler effect which provides unique information of targets is proposed for ground vehicles classification. Firstly, models describing the air-to-ground relationship between ground vehicles and the UAV are built to derive mathematical expressions of radar echo signals. Secondly, Singular Value Decomposition (SVD) is utilized to analyze micro-Doppler components of the ground wheeled vehicle and the ground tracked vehicle respectively. Thirdly, according to classify requirements, specific signal components are restored by Singular Value Reconstruction (SVR). Seven micro-Doppler features are extracted from reconstructed signals. At last, these features are sent to Support Vector Machine (SVM) classifier. Compared with current related methods such as Empirical Mode Decomposition (EMD) and multi-level wavelet decomposition (MWD), experimental results in different cases illustrate the effectiveness of this method. Discrimination performance under various signal-to-noise ratios (SNRs) also proves the robustness of proposed method.