A two-step maximum likelihood algorithm for ISAR motion compensation

Processing of real data has shown that a maximum likelihood approach to translational motion estimation (TME) in ISAR imaging represents a robust method that produces images of fairly good quality. Also, if a persistent scatterer can be tracked throughout an observation interval, a prominent point processing (PPP) method is likely to produce ISAR images of even better quality. We extend the maximum likelihood approach to motion compensation by applying the method again on a local region of the image produced by the first iteration. In some cases the selection of a local region can make it easier to track a persistent scatterer, and a prominent point processing method can also be applied. Radar data of a small aircraft is collected and processed in short intervals, and ISAR images are produced with motion compensation inputs from a combination of processing methods, including phase gradient autofocus. Several hundred ISAR images are produced, and it is shown that a two-step algorithm, with either TME or PPP as the second step, could improve the results in terms of image sharpness when compared to the initial step.