基于异构多核原型芯片的NCS算法并行化

Parallelization of NCS Algorithm Based on Heterogeneous Multi-core Prototype Chip

摘要: NCS算法（nonlinear chirp scaling，非线性调频变标）可以处理大耦合SAR（Synthetic Aperture Radar）回波，实现精确聚焦，但串行NCS算法的成像时间很难达到实时成像要求.为了提高算法效率，采用子孔径结构的NCS改进算法，在自主设计的NoC（Network on Chip）异构多核原型芯片上并行实现了实时NCS成像算法.与串行算法相比，并行化后可以大大缩短成像时间，通过与单次子孔径的理论计算值对比，得出实际并行效率达到90.06%.

Abstract: NCS algorithm (nonlinear chirp scaling) can deal with the large coupling SAR echo,achieve precise focus.However,the imaging time of the serial NCS algorithm could hardly achieve the requirements of real-time imaging.To improve efficiency of the algorithm,we use the improving algorithm with the sub-aperture in this paper.Based on the NoC heterogeneous multi-core prototype chip,we achieve the parallelization of the NCS real-time imaging algorithm.Compared with the serial algorithm,parallel algorithm can greatly shorten the imaging time.And compared with the theoretical calculation value of a single sub-aperture operation,the actual parallel efficiency can reach 90.06%.