1212 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 47, NO. 4, APRIL 2009 Phenomenological Vessel Scattering Study Based on Simulated Inverse SAR Imagery Gerard Margarit, Member, IEEE, Jordi J. Mallorquí, Member, IEEE, Joaquim Fortuny-Guasch, Senior Member, IEEE, and Carlos López-Martínez, Member, IEEE Abstract—This paper presents a study on the origin of the dominating scattering mechanisms observed in polarimetric syn- thetic aperture radar (SAR) images of ships. The study has been made by using numerical simulations, which have been carried out with a radar cross section (RCS) prediction tool (GRaphical Electromagnetic COmputing) and a SAR simulator. Extensive series of simulations has been run for realistic 3-D geometrical models of ships with various sizes. Different radar parameters, aspect angles, and sea surface states have been considered in the scenario. Data analysis with coherent target decompositions has indicated characteristic polarimetric signatures for particular ships within a specific range of viewing angles. This happens at highly oblique incidences where the responses appear to be less sensitive to changes in the operating frequency and bearing angles. Under such conditions, ship scattering can be schematized by the distribution of a set of guide scatterers with high RCS. Their positions and polarimetric characteristics are quantitatively summarized in a new feature vector, which has been proposed to be the basis for classification algorithms. Key ideas about this vector are presented at the end of this paper, jointly with some examples related to three different ships. Recent publications have shown that they can be successfully cast within a new unsupervised vessel classification scheme. Index Terms—Coherent target decompositions (CTDs), polari- metric synthetic aperture radar (SAR) interferometry, SAR po- larimetry, vessel identification. I. I NTRODUCTION I N THE LAST years, fishery authorities worldwide have become aware of the necessity to reach a sustainable ex- ploitation. The increasing world population with a growing consumption per capita is putting intense pressure on ocean coastal areas, overconsuming ocean resources and endangering a complex and fragile ecosystem. One of the main hot spots is overfishing, as recent studies indicate that 80% of the world’s fisheries are nowadays close to their maximum sustainable Manuscript received October 19 2007; revised March 12, 2008, July 25, 2008, and September 23, 2008. Current version published March 27, 2009. This work was supported by the Spanish MEC and EU FEDER funds under Project TEC2005-06863-C02-01. G. Margarit was with the Remote Sensing Laboratory, Department of Signal Theory and Communications, Universitat Politècnica de Catalunya, 08034 Barcelona, Spain. He is now with the Science and Earth Observation Processing Systems Unit, GMV Aerospace and Defense, 08006 Barcelona, Spain. J. J. Mallorquí and C. López-Martínez are with the Remote Sensing Laboratory, Department of Signal Theory and Communications, Universitat Politècnica de Catalunya, 08034 Barcelona, Spain. J. Fortuny-Guasch is with the Institute for the Protection and Security of the Citizen, European Commission Joint Research Centre, 21027 Ispra, Italy. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TGRS.2008.2008443 production [1], [2]. In this context, the enforcement of a tighter fisheries regulation is presently under discussion. The surveillance of the activity of fishing vessels is a very important aspect in this regulation. Current vessel monitor- ing techniques include active transponders [3], optical remote sensing [4], passive acoustic sensors [5], or inverse synthetic aperture radar (ISAR) imagery [6]. The lessons learned in real scenarios have shown that none of these techniques, alone, is sufficiently effective. The main limitations are the high influence of the observation conditions and the lack of tracking autonomy and independence. Other solutions should be then considered, and SAR imagery seems the best option as it can monitor large areas with up to meter resolutions. SAR imagery has been proposed for supporting ship moni- toring (see Improving fisheries Monitoring through integrating Passive and Active Satellite-based Technologies [4], DEtection, Classification and Identification of Marine traffic from Space [7], and Land and sea Monitoring for Environment and Se- curity [8] projects). Some studies have proven its usefulness in ship detection (see [9]–[13] and the excellent review of [14]); however, the studies have also shown the difficulties on classifying vessels by means of their reflectivity properties. Until now, most of the classification schemes are specifically conceived for single-polarization SAR images. They are based on features directly linked to the statistics of the radar ship sig- nature and provide a decision by correlating the measurements with large databases of historical reports [15], [16]. However, despite their established use, their operational exploitation has appeared to be limited because additional information channels are required. Recently, some studies have shown the improved classifica- tion performance that may be obtained with SAR polarimetry [17], [18]. The possibility of isolating more and different features within the ship signature seems advantageous for increasing the discrimination capability. However, the analy- sis of real and simulated data has shown that the reached improvements are still insufficient. Two main restrictions are observed, namely: 1) The resolution of most orbital SAR mis- sions was not high enough for vessel discrimination, and 2) the information about the polarimetric phenomenology of vessels is very limited, and this prevents one from knowing how to interpret and exploit the scattering information for classifica- tion. In the past, this misinformation has caused the extended (but erroneous) assumption that vessel classification cannot be significantly improved by using polarimetric SAR (due to the supposed high sensitivity with respect to the observation condi- tions). This issue is deeply analyzed in this paper, showing the 0196-2892/$25.00 © 2009 IEEE