Accelerated Algorithm for Back-Projection Fully Focused SAR Altimetry
Year: 2025
Type: article
Abstract: Fully focused (FF) synthetic aperture radar (SAR) back-projection algorithms for radar altimetry have gained attention in the altimetry community due to their improved azimuth resolution compared to traditional methods, such as the delay/Doppler algorithm. However, the significant computational cost of back-projection has limited its use in operational applications. This article presents an accelerated back-projection algorithm that reduces the runtime of the classic back-projection by a factor of 28 on a CPU-based architecture. When implemented on a GPU-based architecture, the accelerated back-projection achieves a speedup ranging from 8 to 13 times compared to the classic back-projection on the same GPU. In comparison to the classic back-projection running on a CPU, the combination of the accelerated back-projection and GPU processing results in an overall improvement of up to 1570 times. The accelerated algorithm maintains the accuracy of classic time-domain algorithms and allows for the processing of data in the order of the sensing time under a GPU-based architecture. The algorithm has been extensively validated using transponder and open ocean data. These results indicate that operational back-projection algorithms are feasible for current radar altimetry missions, such as Sentinel-6, and upcoming missions like CRISTAL and Sentinel-3 Next Generation.
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
Award: G1892451603
OpenAlex ID: W4409014483
