Parallel Algorithm Research Of Inverse Synthetic Aperture Radar Imaging

Based on radar characteristic of all-weather,all-day,long distance and wide belt and the ability of easy to distinguish moving target from the fixed background,radar imaging received wide attention.In radar imaging,Inverse Synthetic Aperture Radar can achieve more information of target,which make it have important application value in military science and radar astronomy.ISAR application important value is real-time imaging.However,because of large amount of data and long execution time of the program,real-time imaging of the original pulse echo is impeded.In order to increase radar imaging speed,the primitive means is to develop high performance digital signal processor,but with long development period and high cost.Then,with the prosperity of high-performance technology,parallel computing became the direction of ISAR imaging real-time processing.In allusion to ISAR one-dimensional imaging,in this paper,compressing original echo in distance based on Fourier transform produces one-dimensional imaging.Then by using accumulated cross-correlation method with rectangle window and minimum entropy based on Fourier transform frequency shift properties,one-dimensional range profile is aligned.It is said that using accumulated cross correlation method obtain the coarse aligned range,then based on each coarse aligned profile,multiple offset envelopes is acquired.To calculate the minimum entropy of the offset envelops,the corresponding range of minimum entropy is the aligned one-dimensional imaging of the corresponding pulse.For the large amount of data,high complexity,long operation time and other shortcoming of one-dimensional imaging,based on data parallelism principle of the parallel computing divisions,the paper put forwards ISAR one-dimensional parallel imaging method which applies to multicore processors.By using #pragma omp for of OpenMP instruction,distance compression algorithm and range alignment algorithm is processing parallel.Simulation experiments prove that the parallel method decrease execution time of imaging,and the image quality remain the same.However,the range alignment method based on data parallelism is of low-efficiency.Therefore,based on the new parallel partition pattern which is task parallelism,study the cross correlation parallel method,then combing with minimum entropy data parallel method excutes range alignment hybrid parallel processing.The method using OpenMP instruction #pragma omp section execute parallel processing.Experiment prove that the parallel efficiency is high.However,the new problem is appeared,which is the quality reduction of one-dimensional imaging.In order to ensure the quality of range alignment,in combining with minimum entropy parallel method and cross correlation parallel algorithm which are all based on task parallelism principle,range alignment parallel algorithm is researched.Experiment proved that the method is of high efficiency and the range alignment quality is of guarantee.