Salvatore Maresca

Vessel detection and classification: An integrated maritime surveillance system in the Tyrrhenian sea

In recent years a number of organizations, both national and international, have put significant efforts in developing knowledge-based integrated maritime surveillance (IMS) systems. The final aim is to have a clear picture of the position, classification, identification and movement of cooperative and non-cooperative targets entering and leaving the 200 nautical miles limit of the Exclusive Economic Zone (EEZ). Each sensor (i.e. satellite-based, ground-based, shipborne or airborne) has its own task and, in such a context, high frequency (HF) surface wave (SW) radars are inexpensive tools for long range early warning applications in open waters. They allow maximizing the effectiveness in dealing with fisheries protection, drug interdiction, illegal immigration, terrorist threats, search and rescue tasks. This paper focuses on the possibility of combining automatic identification system (AIS) data with HFSWR data for vessel detection and classification purposes. Three algorithms for target detection in compound Gaussian HF sea clutter are presented and their performance evaluated. The combined use of AIS plots provided by cooperative targets can allow the operator to discriminate non-cooperative targets and possible threats. The concurrent exploitation of AIS and HFSWR data is presented and discussed by means of real data recorded during the NURC experiment in the northern Tyrrhenian Sea in May 2009.

The HF surface wave radar WERA. Part I: Statistical analysis of recorded data

Surface wave (SW) over-the-horizon (OTH) radars are not only widely used for ocean remote sensing, but they can also be exploited in integrated maritime surveillance systems. This paper represents the first part of the description of the statistical and spectral analysis performed on sea backscattered signals recorded by the oceanographic WEllen RAdar (WERA) system. Data were collected on May 13th 2008 in the Bay of Brest, France. The data statistical analysis, after beamforming, shows that for near range cells the signal amplitude fits well the Rayleigh distribution, while for far cells the data show a more pronounced heavy-tailed behavior. The causes can be traced in man-made (i.e. radio communications) and/or natural (i.e. reflections of the transmitted signal through the ionosphere layers, meteor trails) interferences.

Detection, tracking and fusion of multiple HFSW radars for ship traffic surveillance: Experimental performance assessment

Low-power HF surface-wave radars fit well the role of long-range early-warning tools in maritime situational awareness applications, by virtue of their over-the-horizon coverage capability and continuous-time mode of operation. In fact, these sensors, developed for ocean remote sensing, can represent also a further low-cost source of information for ship detection and tracking. Unfortunately, many shortcomings, like poor range and azimuth resolution, high non-linearity and significant presence of clutter, may degrade their performance. In this paper, multi-target tracking and data fusion techniques are applied to experimental data collected during the NATO Battlespace Preparation 2009 HF-radar campaign, which took place between May and December 2009 in the Mediterranean Sea. The system performance is defined in terms of time-on-target, false alarm rate and accuracy. Experimental results are presented and discussed.

The HF surface wave radar WERA. Part II: Spectral analysis of recorded data

This paper covers the second part of the analysis of data recorded by the surface wave (SW) over-the-horizon (OTH) WEllen RAdar (WERA). Data were collected by two WERA systems, on May 13th 2008, during the NURC experiment in the Bay of Brest, France. The principal aim of this work is to provide an accurate characterization of the spectral components of the received signal. Secondly, this information is exploited in order to provide a simple and reliable spectral modeling tool. For this reason, auto-regressive (AR) models, also known as linear prediction (LP) models have been investigated. Our results show that at long distances, when the clutter-to-noise power ratio (CNR) is small, the main components of the spectrum can be reasonably described by an AR(12) model, with a good compromise between accuracy and simplicity. As the CNR increases higher-orders are instead to be preferred.

Radar tracking of a move-stop-move maneuvering target in clutter

In this paper we compare three different sequential estimation algorithms for tracking a single move-stop-move radar target in clutter. We consider optimal and suboptimal Bayesian estimation algorithms, with a special focus on particle filters (PF). The target is modeled using Markov Chains switching theory. Target maneuvers are defined by four different motion models: a stopped target model, a constant velocity model, an acceleration and a deceleration model. We analyze a realistic car traffic scenario by splitting the problem into two study cases. In the first case measurements are expressed in Cartesian coordinates, while in the second we address the problem of nonlinearity in the measurement model. Both cases are characterized by the presence of additive Gaussian noise and by a detection probability less than unity. In addition we are also interested in false measurements originated by high level clutter. The aim of this paper is to compare the so called IMM-PDA-ABF (interacting multiple model, probabilistic data association, auxiliary bootstrap filter) to the well-established Kalman-based PDAF (probabilistic data association filter) and IMM-PDAF (interacting multiple model, probabilistic data association filter) tracking algorithms. Parametric and non-parametric sequential estimation procedures are also taken into account. Advantages and disadvantages of the proposed algorithms are illustrated and discussed through computer simulations.