Time-varying filters for ISAR motion compensation

An approach is presented for the design of time-varying filters that separate selected ISAR target signatures with a constant false alarm rate. Here, the time-frequency representation of each signature is utilized to define a binary filter which is enhanced and labeled using computer vision techniques. Upon filtering, the signatures of the targets are synthesized from their Short Time Fourier Transform representations. In order to obtain a focused radar image for each target, its filtered signature must be motion compensated in the frequency domain. Therefore, radar images can be generated for scenarios in which multiple moving targets can not be imaged otherwise. As an instance, the response of a stepped frequency radar system is simulated to demonstrate that combined signatures of aircraft are effectively separated, compensated and processed into individual-images.