Skip to Main Content
When Doppler centroid estimators are applied to satellite synthetic aperture radar (SAR) data, biased estimates are often obtained because of anomalies in the received data. Typical anomalies include areas of low SNR, strong discrete targets, and radiometric discontinuities. In this paper, a new method of Doppler centroid estimation is presented that takes advantage of principles such as spatial diversity, estimator quality checks, geometric models, and the fitting of a "global" estimate over a wide area of a SAR scene. In the proposed scheme, Doppler estimates are made over small blocks of data covering a whole frame, so that all parts of the scene are potentially represented. The quality of each block estimate is examined using data statistics or estimator quality measures. Poor estimates are rejected, and the remaining estimates are used to fit a surface model of the Doppler centroid versus the range and azimuth extent of the scene. A physical model that relates the satellite's orbit, attitude, and beam-pointing-direction to the Doppler centroid is used to get realistic surface fits and to reduce the complexity (dimensionality) of the estimation problem. The method is tested with RADARSAT-1 and Shuttle Radar Topography Mission X-band SAR (SRTM/X-SAR) spaceborne data and is found to work well with scenes that do have radiometric anomalies, and in scenes where attitude adjustments cause the Doppler to change rapidly.