ABSTRACT
Inverse synthetic aperture radar (ISAR) has been studied for more than three decades and several demonstrations have proven its effectiveness in forming electromagnetic images of non-cooperative targets [1] [20]. In its most simple form, ISAR systems produce 2D images of targets. A 2D-ISAR image can be interpreted as a filtered projection of a 3D target’s reflectivity function onto the image plane. Given the dependence of the image plane orientation on the radar-target geometry and dynamics (which are typically unknown), such projection cannot be predicted. This often results in a difficult interpretation of an ISAR image. The lack of knowledge of the projection of the target onto the image plane necessarily causes difficulties in the classification or recognition of targets by using ISAR images. Although there have been some attempts
to estimate the orientation of the image projection plane [15], which directly relates to the estimation of the effective rotation vector, the applicability and the effectiveness of such techniques are yet to be proven. A radical solution to this problem is to form 3D ISAR images, which completely eliminates the problem of dealing with an unknown projection. Previous attempts to form 3D ISAR images can be found in the literature. One of such approaches aims at forming 3D ISAR images by exploiting single sensor ISAR image sequences [6] [17]. 3D target motions, in fact, produce a set of view angles that allow for the estimation of the 3D position of each target scattering center. This approach has the advantage of requiring a single sensor, although it relies on long target observation time intervals and on the a priori knowledge of the target’s motions [14]. Another approach is based on interferometric principles and makes use of multiple sensors [19] [2] [8] [21]. Such an approach has the advantage of not requiring long observation time intervals nor the a priori knowledge of the target motions. Classic interferometric techniques uses range profiles and are based on the assumption that a single scatterer is present in a resolution cell. Therefore, the probability of distinguishing more than one scatterer in a range resolution cell is much lower than in the case of a 2D (range and Doppler) resolution cell, obtained when using 2D ISAR imaging. The layover effects are then minimized due to the higher probability of discriminating more than one scatterer in a single range resolution cell. As a consequence, the estimation of a scatterer’s position along the cross-range might be improved.
