SAR Doppler center frequency parameters estimation parallel optimization based on multi-core ARM NEON technology

摘要

Obtaining Doppler center frequency parameters estimation (DCFPE) with low time consumption is a great challenge in synthetic aperture radar (SAR) imaging system. In order to get DCFPE in real time, this paper adopts three parallel acceleration implementation methods-SIMD parallelism, multi-issue parallelism and multi-core parallelism-according to the characteristics of DCFPE algorithm. To verify the efficiency of the proposed methods, the paper builds the ARM embedded Linux system on multi-core ARM Cortex-A9 platform and uses assembly programming language to implement DCFPE algorithm in a 16384*16384 64-bit SAR imaging process. The experiment shows that DCFPE can be accelerated by more than 50% on average with an acceptable accuracy of the result.