Direct image formation with current distributions generated by shooting and bouncing rays

摘要

Image formation by inverse synthetic aperture radar (ISAR) methods is one of the most advanced approaches to explore the scattering or radiation properties of a finite sized object. ISAR imaging is based on the coherent processing of radar signals, which are collected for a range of observation angles and for a certain range of frequencies. In a radar experiment, it is mandatory that ISAR works with the waves scattered from the observed object. In contrast, in simulation based considerations there is no need to compute the scattered waves explicitly. It is rather recommended to directly generate the ISAR image with the induced currents on the targets, which are usually available in an electromagnetic simulation, e.g., by the shooting and bouncing rays (SBR) techniques utilizing physical optics (PO). Instead of computing the scattered or radiated fields from the real or equivalent currents the radiation integral is directly inserted into the imaging integral and by interchanging the integration orders, the imaging point spread function can be generated. Consequently, the image formation is reduced to a convolution of the found point spread function with the current distribution. A concise vectorial formulation of this well-known methodology is presented together with a discussion of important properties. The general case of 3-D ISAR imaging is considered, which is also specialized to the 2-D situation. The point spread functions are analytically derived for narrow angle and narrow bandwidth imaging, where a bistatic observation range symmetrically arranged around one incident direction is considered. The resulting images can thus be assumed as a good approximation of monostatic images, which are often desired. Various examples of 2-D and 3-D images for complex metallic objects such as automobiles are shown, which have been obtained from the surface currents of an SBR field solver. Implementation issues related to the required interpolations as well as the efficient r--ealization of the SBR simulations are discussed.