An Improved Doppler Parameter Estimator for Synthetic Aperture Radar

Abstract

To achieve a high quality synthetic aperture radar (SAR) image, the motion caused by the radar and/or the moving target needs be compensated for coherent processing, and this requires the accurate estimation of the Doppler parameters of the received signals. A new, ap- proximate, maximum likelihood (ML) estimator in the time domain is obtained recently to jointly estimate the Doppler parameters, and it is shown to have much better performance than that of an existing approximate ML estimator in the literature, for the values of the Doppler parameters of interest. The efiects of the parameter estimation errors on the SAR system performance are analyzed. By revealing the relationship between the range/azimuth resolution of SAR imaging and the estimation accuracy of the Doppler parameters, we show that our new estimator can be applied in SAR imaging to improve the image quality. 1. INTRODUCTION With the reputation of high resolution and impressive quality of image, synthetic aperture radar (SAR) has played an important role in cartography, oceanography, and numerous military applica- tions (1). Since the signal energy from a point target is spread in range and azimuth, the purpose of SAR focussing is to collect this dispersed energy into a single pixel in the output image. The optimum focusing of the SAR data is a space-variant and two-dimensional operation, which makes SAR processing a challenge. The most popular SAR processing algorithm is the Range-Doppler technique (2) and its variations; see (3) and (4). The method is e-cient, and in principle, solves the problems of azimuth focussing and range cell migration correction. These imaging algorithms require the accurate estimation of the Doppler parameters, namely, the centroid Doppler frequency and the frequency rate, to perform coherent processing. Our main goal here is to estimate accu- rately the Doppler parameters to compensate for the motion caused by the radar or the moving targets. This compensation is challenging and important for moving targets, where the motion is non-cooperative as in inverse SAR (ISAR) (5). Without correct motion compensation, the im- age quality may be degraded in several ways, such as shifting, distortion, defocusing and so on. The traditional estimation for the centroid Doppler frequency and the frequency rate is usually per- formed separately for simplicity by using clutter-locking and autofocus technologies, respectively (6). However, this leads to error propagation, and thus, the optimal estimator is to jointly implement clutter-locking and autofocus. The conventional joint estimation in the frequency domain (7) is a two-dimensional, nonlinear search. Unfortunately, there is still no exact, closed-form solution for solving the general nonlinear programming problem, and the search complexity is high. In this paper, we present a new, approximate, maximum likelihood (ML) estimator in the time domain. It shows that for the range of values of interest of the Doppler parameters (6), our estimator leads to a better performance than that of the only other existing approximate, ML estimator in the time domain, i.e., the Djuric-Kay (DK) estimator (8).