Design and Implementation of Doppler Centroid Estimation with Quality Index for Real-Time SAR Imaging

Doppler centroid estimation is essential for synthetic aperture radar (SAR) imaging in stripmap mode. However, the quality of estimation is influenced by the scene significantly. We propose to use dispersion of phase differences from beat signals of adjacent range lines as a quality index for selecting reliable estimates. The improvement in success rate demonstrates its effectiveness compared to the conventional quality indices such as the signal-to-noise ratio (SNR) and contrast. A hardware accelerator for real-time Doppler parameter estimation including baseband Doppler centroid and Doppler ambiguity estimations is then implemented. It is integrated in the range Doppler algorithm (RDA) imaging processor and supports configurability. The range compressed data in the frequency domain are used as the input. Several hardware design techniques are adopted to reduce buffer sizes, computation complexity, and latency. The processing time of our Doppler centroid estimator for an $8 \mathrm{~K} \times 8 \mathrm{~K}$ image is 0.082 s at 100 MHz operating frequency.