Accurate parameter estimation using a frequency domain maximum likelihood approach for SAR processing

A frequency domain maximum likelihood (ML) approach is proposed to estimate Doppler parameters of Synthetic Aperture Radars (SAR). The estimator includes Doppler centroid and Doppler frequency rate. Radar echoes are modeled as a zero mean complex Gaussian process. The range Doppler spectrum is obtained by applying azimuth Fast Fourier Transform (FFT) to raw data without range compression. It is derived that the average range Doppler spectrum is a Gaussian process with its mean depending on Doppler parameters and its unknown covariance. Thanks to the prior knowledge of antenna radiation pattern, it is possible to estimate the covariance of the spectrum noise. As a consequence, maximizing the likelihood may reduce to a weighted nonlinear least square optimization issue. The parameters can be reliably estimated by using two-dimensional search. A RELAX based iterative algorithm is applied to accelerate the craft search. Both computer simulation and experiment data demonstrate that the frequency domain ML approach may provide accurate Doppler parameters for SAR processing.