Implementation Technology Of Array Radar Signal Processing Based On GPU

Nowadays,array radar is a kind of high-performance radar with good application prospects.This kind of radar has many advantages: multi-target,long-range,high data rate,high reliability,high adaptive ability,fast and accurate beam conversion ability,and can interleave in time or even complete many functions simultaneously.In recent years,GPU technology has been developed rapidly.Software radar based on GPU has become a hotspot of current research.Compared with the traditional hardware radar based on FPGA + DSP,the software radar has the advantages of low development cost,short development cycle,easy debugging,real-time and strong expansibility.Supported by a certain type of phased array radar system,this thesis focuses on the implementation technology of array radar signal processing based on GPU.The main work includes the following aspects:Firstly,the heterogeneous architecture of CPU + GPU is introduced,and the programming model of CUDA is analyzed in detail,which includes the execution process,kernel function,thread hierarchy and storage structure of CUDA program.Then,the signal processing flow of the radar test system is introduced,and the key algorithms in array radar signal processing are studied,which includes digital beamforming,pulse compression,moving target detection,constant false alarm detection and angle measurement.Moreover,a method to eliminate false alarm points of sidelobe targets based on array processing is proposed.The new method can reduce the false alarm rate and improve the detection performance of targets,which can also be used to solve the problem of high false alarm rate caused by high sidelobes in conformal array.Next,according to the principles of signal processing algorithm,the actual received radar echo data are simulated and analyzed and the processing effect of each algorithm is further explained.Finally,the implementation of array radar signal processing algorithm on GPU is studied.Firstly,the execution time of signal processing algorithm under GPU and CPU is compared.From the acceleration ratio,it can be seen that signal processing under NVIDIA Tesla K40c GPU can fully meet the real-time requirements.Then,the realization principle of GPU-based software radar and the serial-parallel allocation of program are analyzed in detail.The implementation process of digital beamforming,pulse compression,moving target detection and CFAR detection on GPU is studied in depth.Furthermore,combined with the actual echo data received by radar,the operation results of PC algorithm,MTD algorithm and CFAR algorithm under Matlab and GPU are compared respectively.From the difference chart,it can be seen that the processing results under the two platforms are basically identical.GPU-based array radar signal processing technology studied in this thesis can be well applied to the development of software radar system.