Parallel/pipeline multiprocessor architectures for SAR data processing

Abstract

In SAR digital data processing, using the range‐Doppler approach, the main computational load occurs in the range and azimuth compressions, for which one‐dimensional circular convolutions in the Fourier domain are used. In the paper it is shown how the intrinsic degrees of parallelism of the process can be efficiently exploited to design a flexible parallel/pipeline multiprocessor architecture, based on advanced components with interconnection capabilities and capable to perform up to the real‐time SAR data focusing. In the proposed architecture, an efficient line compression can be achieved by means of pipelined processors for the range and the azimuth compressions and many of these structures, each performing the compression of a line or its section, can be used to sustain the computational load and the data rate requirements. In the paper, we also show how the data communication between the range and the azimuth processors can be solved when a maximum of eight of these line compressors are used. The performance evaluation of the proposed architecture is made referring to real spaceborne SAR sensors.