Research On The Fpga Design Of Large Array Imaging Algorithm With OpenCL

Large antenna arrays imaging radar is a new kind of radar system developed in decades.It has a characteristic of high azimuth resolution,so that we can get clear target images from it.this kind of radar,with large array imaging technology,can get a clear radar image without a relative deformation between the antenna arrays and the target.Nevertheless,the data handling have great capacity is to makes large antenna arrays imaging radar a little bit difficult.It needs a high-performance data processing device to meet the algorithm's real-time requirement.OpenCL(Open Computing Language)standard provides a new perspective and direction to the study and settlement of the real-time problems.As one kind of heterogeneous computing standard,OpenCL enables large array imaging algorithm that benefit from parallelization of data processing.This paper is about the large array imaging algorithm implementation with OpenCL based on FPGA platform,and it contains the following parts:1.Learning the technology and algorithm of large array imaging.Completing the modeling and simulation of large array imaging algorithm.Verifing the performance of the algorithm program in MATLAB environment and then converting the MATLAB program into C language program.2.Parallelizing the serial large array imaging algorithm to parallel algorithm and accomplishing OpenCL programming tasks.Owing to its particularity,programming the OpenCL program that run on the GPU first to prove reasonable and efficient of the parallel algorithm.Completing the OpenCL program that run on the FPGA according to the speciality of FPGA at last.It's important to note,however,that the OpenCL program was divided into multiple sections because of FPGA resource constraints.For this reason,the large array imaging OpenCL programs running on FPGA were running separately.3.Optimizing the OpenCL program running on the FPGA.Due to the hardware resource constraints,it's unrealistic to optimizes the whole program thoroughly.Optimizing the divided programs within the FPGA platform resource constraints instead of the whole program and analyzing its performances.4.Comparing and analyzing the output data from all the programs above given the same input simulation data.The results are quite encouraging in that the OpenCL programs show good speedups although the programs is seriously restricted in the aspects of FPGA hardware resources.