Research And Implementation Of Video SAR Echo Simulation Based On GPU

Video Synthetic Aperture Radar is an active microwave remote sensing system.Compared with sensors such as optical and infrared sensors,the video synthetic aperture radar has strong cloud,smoke,and shallow surface penetrating ability and can achieve all-day,all-weather and high-resolution imaging.Compared with the traditional SAR,the video SAR can perform continuous multi-frame observation and imaging of the ground scene target,and achieve speed-independent detection of the moving target.It belongs to the frontier technology of radar research.In the development process of video SAR,it is necessary to use echo data for imaging algorithms,motion compensation algorithm verification,parameter setting optimization,and system performance evaluation.Due to the high acquisition cost of the measured echo data and the high technical requirements,the echo data obtained by simulation is favored by researchers.However,in the face of the demand for high frame rate,high resolution and wide swath video imaging of video SAR,the amount of data and the simulation time for echo simulation for distributed scene targets have increased dramatically.In recent years,research on parallel simulation based on graphics processors has gradually deepened.Based on the above background,this paper focuses on the simulation of video SAR echo based on GPU.Firstly,the echo signal model is deduced by introducing the synthetic aperture principle.By separating the transmitted signal and the system impulse response function in the frequency domain,an echo model based on a one-dimensional frequency domain Fourier transform is obtained and combined with video SAR applications.Analysis of the one-dimensional frequency domain Fourier transform method reduces the echo accuracy by shifting the main lobe of the signal during pulse coherent accumulation,and then introduces a sinc function to improve the echo amplitude characteristics.By analyzing the signals and performing pulse coherent accumulation according to their own distance units,it is concluded that the spacing of the ground scattering points should be the ground projection from the sampling interval.Analysis of the traditional methods for calculating beam coverage by calculating the range and slant angles of all the scattering points of the scene has great computational redundancy.A method for screening beam coverage based on radar imaging geometry is proposed.In addition,according to the actual work scene of the video SAR,a signal model of stripmap and spotlight is constructed,and a complex scene echo simulation method including moving targets is proposed.Secondly,through introducing the heterogeneous programming mode of CUDA based on "CPU + GPU",analyzing the CUDA main program of CPU side to carry on the logic processing and controlling the way of GPU running,and analyzing the CUDA kernel program loading into GPU stream multi-processor in the form of thread block.By introducing the storage capacity,access speed,and usage of different types of memory on the GPU and the way in which the Fourier transform is implemented in CUFFT in parallel,the parallel implementation technology foundation for the GPU-based video SAR echo simulation is provided.The parallel analysis of video SAR echo simulation and the design of parallel processing framework are completed.The parallel processing is improved in four aspects: task block processing,pulse coherent accumulation optimization,merge memory access optimization and channel transmission optimization.Through parallel optimization,the efficiency of simulation of large-area high-resolution video SAR echo signals is improved.Finally,through simulation experiments,it is verified that the improved echo signal model can improve the echo signal amplitude.Grouping experiments were performed to verify the accuracy of video SAR echo simulation under various imaging modes and complex scenes containing moving targets.Echo simulation time statistics for scene objects of different sizes were performed.The experimental results show that the GPU-based video SAR echo simulation method described in this paper has greater efficiency.