Accelerating ocean currents analysis: tiling-based parallel computing for SAR imagery

Abstract

High-resolution synthetic aperture radar (SAR) data processing for marine applications is computationally demanding due to large memory requirements. A novel parallel approach for SAR image clustering is introduced, leveraging MATLAB’s distributed computing to achieve fast and precise SAR image segmentation. We make use of this efficient parallel computing framework for Doppler centroid estimation (DCE) using the correlation Doppler estimator (CDE) and sign Doppler estimator (SDE). An onboard tiling method partitions the SAR scene into tiles of blocks, enabling localized estimation of the baseband Doppler centroid component fDC. These estimates are unwrapped across range and azimuth while preserving resolution. Rigorous quality control ensures accuracy, and Doppler signatures are analyzed to derive ocean surface currents (OSC). The proposed parallelized approach reduces computational load by 32 times, enabling near-real-time OSC estimation for operational SARbased ocean monitoring. Moreover, statistical indicators including bias, correlation, and slope are promising and validate the effectiveness of parallel computed coarse data for marine applications