1 Validation of Complex Naval Target Models using Superresolution Imagery Methods F. Totir, METRA Research Institute, Bucharest, Romania E. Radoi, A. Quinquis, E 3 I 2 Research Centre (EA 3876), 2 rue François Verny, Brest, France Abstract: The main purpose of the paper is to propose a framework for validating radar target models using superresolution imagery methods. The need for realistic target models becomes more and more imperative as the radar technological capabilities and the signal processing techniques are continually improved. Simple scattering configurations, such as a set of well resolved scattering centers, where no geometrical visibility or secondary reflections are taken into account, are going to become completely inappropriate. A complex target model is considered in the paper, the simulated backscattered signal being obtained via the “simplest component method”. The target’s body is described by a set of elementary geometrical surfaces, whose reflection characteristics are known. Superposing partially scattered signals provides the global echoed signal. This way, a direct and natural link with ship’s physical structure is implemented. Geometrical visibility but also interferences between components are also considered. A superresolution technique (MUSIC-2D) is then used, in order to validate the proposed model. I. INTRODUCTION One of the most stringent requirements of development and implementation of radar target classification systems is to have a large database of radar signatures. This is imposed by classification tests to be performed but also in order to validate the image reconstruction algorithms and target synthetic models. Synthetic target signatures are widely used in algorithms development phase, as it is difficult and expensive to build a real target signature database. However, usual target models tend to be rather simplistic (i.e. a set of scattering centers). Moreover, this model is supposed not to change during the target’s rotation (i.e. no partial masking is taken into account) and no multiple reflections are considered. Another weakness of this kind of models is that they are not directly linked to the corresponding target’s physical structure. All these flaws are mainly due to the difficulty to electromagnetically describe complex man-made objects. Realistic models are even more difficult to design for naval targets because of their structure and movement complexity. The sea clutter, which becomes very spiky for high resolution radars and low grazing angles, is another key feature to be taken into account. Radar imagery is an effective tool for characterizing targets with complex structures. Hence, it is also very suitable for evaluating the matching between a given model and the corresponding real target. The target model can be thus validated according to the application requirements. An integration time as short as possible should be considered for the imaging process in order to keep uniform the target’s motion during this period. However, this choice will limit the angular integration interval, so that the achieved transversal resolution is poor when Fourier transform is used for the image reconstruction. Superresolution methods have the advantage to provide high resolution, even for very limited angular domains and frequency bands. These techniques are mainly based on the eigenanalysis of the data covariance matrix and their use is advantageous especially for maneuvering or very mobile targets. II. REALISTIC TARGET MODEL We use the “simplest component method” [1] in order to link the considered scattering center distribution to target’s physical structure. This method combines elementary surfaces and bodies with known scattering characteristics (rectangular, spherical, cylindrical surfaces etc.) to build the target. The global echo is obtained by the superposition of all the signals scattered by individual components. A direct link between target’s physical structure and scattered signal is established this way. Three views of the considered naval target model are given on Figure 1. The inferior part of the target is defined using a set of rectangular and triangular plates. Thus, prove is composed by three triangular plates; the other part is made by assembling four rectangular plates. It is important to notice that edges between these plates are modeled as stand-alone elements, since they are also stand-alone scattering elements. To describe the superstructure of this model, quadratic surfaces are generally used. The two gun turrets are seen as semi-spheres while the command deck and the engine’s evacuation tube are modeled as elliptical cylinders. Figure 1 Target model Each of these components is described using its own local coordinate system, whose orientation and position are also given as input parameters. Because mathematical generated surfaces are often infinite (e.g. cylindrical surface), limiting surfaces (pair of planes) have been used to retain only useful areas. This is also the case when only