Chapter
1
Introduction
1.1
Motivations for research
Continuous monitoring of the maritime national borders, the so called Exclusive Economic Zone (EEZ), takes on great significance, both in terms of ship and aircraft surveillance1. The broad range of requirements for maritime surveillance is intended for supporting the protection and exercise of national sovereignty, not only in terms of law enforcement but also in terms of search and rescue, environmental protection and resource management.
As a matter of fact, intensive research activities have been pursued in order to exploit ex-isting sensor systems in support of maritime surveillance. In this domain, a variety of sensors can be used, including coastal radars, video, infra-red (IR), synthetic aperture radar (SAR) sensors and automatic identification system (AIS) tracks. Sensor integration for surveillance purposes has been widely addressed in the literature, but it is still a challenging issue and it requires deeper studies in order to solve the problems related to the different nature of the sensors, their internal characteristics and the type of information provided. In this sense, some interesting readings are [Bar95], [Cas94], [Cha86], [Deb93], [Far86], [Gio08] and [Mah07]. However, in such a scenario, long-range affordable maritime surveillance systems op-erating on a continuous time basis are essential. Traditional land-based microwave radars are limited to operate within line-of-sight (LOS) propagation. The EEZ can be covered by a number of airborne and ship-borne sensors, but these provide only a small snapshot in time of activity within the EEZ. Sky-wave OTH radars can be used for this purpose, but they need large installations and are too expensive2. Satellites have neither the spatial nor the temporal resolution to provide the necessary level of real-time surveillance.
HFSW radars could be a reliable though inexpensive tool for low/medium level monitor-ing accuracy of long range areas. They grant early warnmonitor-ing pictures of the surveyed zone and, most of all, they are able to fill the gap of spatial coverage represented by microwave radars and do not suffer the periodic temporal coverage of satellite-based sensors.
Further-1Following the United Nations Convention of the Law of the Sea (UNCLOS), coastal nations are required to
establish and maintain Administration, Law Enforcement and Environmental Protection over the 200 nautical miles (nm) of sea, far beyond the 12 nm territorial limits.
2For a complete survey on Sky-wave OTH radars, see [Sko08].
Chapter 1 - Introduction
more, they are simple and economical systems, since the operating costs per unit of area are significantly smaller than those of microwave radar systems. These characteristics make HFSW radars appealing for monitoring and surveillance of both low altitude aircrafts and surface vessels with a coverage span which can extend as far as the EEZ. For this reason, the backbone of integrated maritime surveillance (IMS) systems is generally made up of with HFSWRs [Sev01], [Pon01a] and [Pon01b].
However, despite the aforementioned benefits, HFSW radars suffer from a series of shortcomings3. The first problem to deal with is the extreme crowding of the HF spectrum. Finding a free HF channel for transmitting/receiving is therefore a complicated procedure which involves an accurate scan of the available frequency interval [Gur07]. Once the prob-lem is addressed, sea clutter often mask target-generated echoes in large portions of the range-Doppler (RD) spectrum, thus rendering detection a very challenging problem to deal with [Kha91], [Tur97].
The contribution of HF sea clutter is produced by specific spectral components of the surface-height wavefield. If the ocean wave spectrum contains sea wavelengths of the order of magnitude of the radar wavelength, the Bragg scattering theory is applicable. First-order scattering is due to those ocean waves of half the radar wavelength which travel towards and away from the radar site. The Doppler spectrum of the backscattered signal then contains two lines, corresponding to the phase velocities of the scattered ocean waves. These frequencies often deviate from the theoretically known values in non-moving waters. This phenomenon is attributed to an underlying surface current. Finally, the full ocean wave spectrum is due to the second-order scattering which generates side-bands in the HF Doppler spectrum. The basic physics of backscattering of electromagnetic waves were discovered and described by Crombie [Cro55] and Barrick [Bar72].
In addition, the HF signal is prone to a wide number of interference types, both man-made and natural. External interference from natural and man-made sources typically masks the entire range-Doppler search space and is characterized by a spatial covariance matrix that is time-varying or non-stationary over the coherent processing interval (CPI). This physical phenomenon may arise from a number of causes. For instance, the dynamic properties of the ionospheric layers propagating the HF interference, the variation in geometry between radar receiver and interference sources and the impulsive nature of the sources [Hea74], [Fab00] and [Sko08]. Unoccupied HF channels are very difficult to find, especially at night, when the ionosphere is prone to propagate interference at long distances. The space-time characteriza-tion and modeling of ionospherically propagated signals can be found in [Fab00]. For this reason, adaptive beamforming (ABF) techniques are usually applied in order to suppress such interferences effectively [Fab04].
Regarding ship detection, the task is to resolve targets in the Doppler domain from the same background clutter exploited for sea-state sensing. The two problems are thus in many ways complementary. In fact, the presence of clutter is unwelcome as far as ship detection is concerned, while the presence of ship returns can negatively affect the extraction of oceano-graphic parameters [Gur05]. For this reason, in the past years there has been much interest in trying to develop new spectral models the return from the sea directly, with the ultimate goals of both enhancing target detection via clutter-suppression techniques [Kha91], [Kha94] and sea current sensing [Mar97].
3Here only a few of them are mentioned. For the interested reader we suggest [Sko08]. Though the discussion is
mostly concerned with Skywave OTH radars, under several aspects it can be applied to HFSW radars as well.
1.2 Main contributions
Recently, great attention has been paid to low-power HFSW radars4. The Wellen Radar (WERA), developed at the University of Hamburg is such a system [Gur99a], [Gur99b] and [Gur01]. WERA is commonly used for oceanic applications (e.g. sea current sensing, waves and winds mapping), but thanks to its inexpensiveness and easiness of installation, if compared to common high-power HFSW systems, it could represent a precious OTH early warning tool in IMS systems, as pointed out in [Dzv07], [Dzv08] and [Dzv09b].
1.2
Main contributions
In this new emerging context the main contributions of this work are the following: • Data measured by the two WERA systems have undergone an accurate statistical
analysis, with the final purpose of acquiring knowledge about the characteristics of the signal, both in the time and frequency domains. This analysis has led to the conclusion that sea clutter in the HF can be modeled as a compound Gaussian process. In other words complex Gaussian process (speckle) modulated by a slowly non-negative process (texture) [Gre09a], [Gre09b] and [Mar10c].
• Ship detection has been investigated in non-stationary compound Gaussian sea clutter. By exploiting the results obtained in the first part, knowledge about its distribution has led to the application of the normalized adaptive matched filter (NAMF) test, where the knowledge about clutter, in terms of its covariance matrix, has been evaluated by means of three algorithms, namely the sample covariance matrix (SCM), the nor-malized sample covariance matrix(NSCM) and the fixed point covariance matrix (FPCM), derived from the approximation of the maximum likelihood (AML) estima-tor [Mar10a], [Mar10b].
By considering all these factors and the new emerging interest in reliable and inexpensive radar systems for long range detection, this thesis adds interesting contributes to the overall body of knowledge about HFSW radars.
1.3
Outline
This thesis is focused on the statistical processing of data recorded by two WERA systems, during the NURC experiment in the Bay of Brest (France), and is organized as follows: Part I deals with the statistical and spectral analysis and modeling of the signal, while Part II is concerned with ship detection.
Chapter 2 describes briefly an HFSW radar system. Details about its radar equation and differences with common microwave radars are provided as well. Chapter 3 concentrates on the description of the WERA system and the NURC experiment in the Bay of Brest. Prelimi-nary analysis on the received data is performed and discussed in the final part of the Chapter. Chapter 4 describes an in depth statistical analysis on the measured data, while in Chapter 5 the spectral analysis is carried out instead. Finally, in Chapter 6 a comprehensive study on detection algorithms is presented and results are discussed. Conclusions and guidelines for future work are reported in Chapter 7.
4They are typically exploited for remote sensing applications.
Chapter 1 - Introduction
