Automatic vehicle classification in a low frequency forward scatter micro-radar

摘要

Forward Scattering Radar (FSR) is a special case of bistatic radar configuration where the desired radar signal is formed via the shadowing of the direct (transmitter-to-receiver) signal by the target body. FSR offers a number of interests including an inherent ability to detect stealth target, absence of signal fluctuations, reasonably simple hardware, enhanced target radar cross-section (RCS) compared to traditional radar and capability to use Inverse Synthetic Aperture algorithms for Automatic Target Classification (ATC). Of course as any system FSR has its own drawbacks and limitations. This thesis presents the research results on development of ATC algorithm under a variety of external factors such as clutter and target's trajectories uncertainties. The peculiarity of this research are that the FSR operates at a low (VHF and UHF) frequency bands that in a strict sense does not correspond to an optical region for vehicles like targets and the system operate with omnidirectional antennas. There is no previous research considered this practically important case. The algorithm is developed based on Fourier transform, Principal Component Analysis (PCA) and K-Nearest Neighbour (KNN) classifier - for features extraction, transformation and classification, respectively. The ATC system is integrated with coherent signal processing algorithm in order to estimate target’s motion parameters (i.e speed) prior to spectra normalisation process. The analytical and modelling results are experimentally confirmed. As ATC performance degraded when high level of clutter is present, cluttercompensated ATC model is introduced and its classification performance is analysed using measured signals with added simulated clutter.