Research On Application Of Distributed Computing In Synthetic Aperture Radar Imaging

Synthetic Aperture Radar(SAR)imaging technology has been widely used in reality,however,with the complexity of application scenarios,more and more computation of SAR imaging results in a series of problems,which limits its further application in reality.In view of the fact that distributed computing has been applied in the scenarios that need huge computing,and has realized the fast processing of massive data and the flexible expansion of computing power.This paper introduces the basic principles of Back Projection(BP)algorithm and Compressed Sensing(CS)imaging method of SAR.Based on the parallel computing ability of the MapReduce distributed computing framework,two distributed parallel computing methods of SAR imaging are proposed.1.The BP algorithm has wide applicability in SAR imaging.However,when SAR uses BP algorithm to image,there are problems of large calculation amount,long imaging time,and insufficient scalability of the signal processing platform computing power.To solve these problems,this paper proposes a fast imaging method of SAR back projection based on MapReduce.In this method,the azimuth imaging task of BP algorithm is divided into several imaging units,carry out distributed parallel azimuth imaging calculations.Finally,the calculation results of all imaging units are coherently accumulated.In this method,the digital mark method is used to record the position information of the antenna array element corresponding to each pulse,so as to realize the independent data processing;the combiner function is used to aggregate the calculation results in the imaging unit in advance,so as to solve the problem of long data aggregation time in the later stage.Experiments verify the accuracy and acceleration performance of the method.The azimuth imaging of this method is carried out in a Hadoop distributed computing platform built by 4 physical computers,and its calculation speed is 3.7 times of the azimuth imaging of the BP imaging method of single machine calculation.It can be seen that this method can realize the acceleration of BP imaging.2.In the existing SAR system,in order to reduce the number of sampling points,the CS imaging method is commonly used in the near field,but the SAR adopts this method,which has the problems of long calculation time and insufficient scalability of the signal processing platform computing power.In order to solve the above problems,this paper proposes a fast imaging method of SAR compressed sensing based on MapReduce.In this method,MapReduce computing process is applied twice to realize the iterative reconstruction of range imaging and azimuth imaging orderly by distributed parallel calculation.In this method,the serial number marking method is adopted to realize the distributed parallelized reconstructed data which can be sorted according to the row order before reconstruction.The experimental results verify the imaging accuracy and acceleration performance of this method.When this method is running in a distributed computing cluster with 4 computing nodes,its computing speed is 1.4 times of the SAR compressed sensing imaging method of single machine calculation.It can be seen that the method can realize the acceleration of CS imaging.