Detection of ships using OTH radar with short integration times

Abstract

Localizing and tracking ocean-going ships using high-frequency over-the-horizon radar (OTHR) currently requires long coherent integration times (CITs) and places critical demands on scarce radar resources. Here, modern spectral analysis techniques (MSATs) are employed to localize ships from clutter and noise in Doppler space using shorter CITs than are possible with standard Fourier techniques. Doppler superresolution of the MSAT is achieved by employing statistical knowledge of the signal's structure. The nature of the statistical assumptions is discussed, and potential problems are outlined. The MSAT we have adopted is a variant of that of Tufts and Kumaresan (1982), with an information-theoretic criterion for determining the number of poles represented by the autoregressive model and an additional step to provide accurate SNR estimates suggested to us by Friedlander. We improve on the MSAT even further by adding a preprocessing step, forming a filter to remove the Bragg lines. The algorithms have been tested on data collected both at the California-based Wide Aperture Research Facility (WARF) and the US Navy relocatable OTHR (ROTHR-VA). In both cases, factors of 3 to 8 reduction in CITs are commonly achieved with no loss in resolution. Under somewhat favorable conditions, ships near the Bragg lines have been localized using only 2-second CITs, as used for aircraft detection.