Time-Frequency distributions for SAR doppler rate estimation

The estimation of the best focusing parameters directly from the data is of particular interest in the processing of SAR imagery whenever accurate orbit and attitude information is not available for the satellite-borne sensor. The basic problem is to determine the doppler rate of chirp signals. Time-frequency distributions have been proposed for such application and they prove accurate enough when echoes from strong point scatterers, present within the images area, are used. However, when natural reflectors instead of artificially deployed corner reflectors are used, results are no longer reliable, usually because of the multiplicity of apparently single scatterers. In these cases simple methods will not work, as their outcomes vary too much because of border effects. Hough transform performs well, but it is computationally heavy if accurate estimates are needed. Here the slope of the principal axis of the time-frequency distribution is used to estimate chirp rate. An iterative procedure is used to limit the influence on the accuracy of the results of border effects, background noise and contributions from smaller scatterers at the same slant range. Results are good. A few percent error in the estimated chirp rates has been observed with ERS-1 data.