Research On Radar Emitter Recognition Technology Base On GPU Acceleration

The complexity of electromagnetic environment makes the emitter recognition system face greater challenges.To identify the emitter at high speed can reflect the advanced nature of the reconnaissance system to a certain extent.With the increasing complexity of radar modulation types and the increasing interleaving and overlapping degree of radar pulses,it is difficult for traditional radar signal processing algorithms to process the received signals quickly.Parallel processing of radar signal is one of the effective ways to solve the problem of real-time and accuracy.The development of parallel computing technology provides conditions for radar signal parallel processing.It is necessary to apply GPU to radar signal processing.This thesis uses a general parallel computing platform CUDA,which is a heterogeneous platform based on CPU and GPU,and can give full play to the advantages of CPU and GPU.Based on the background of high-density electromagnetic environment,this thesis focuses on GPU accelerated radar emitter recognition technology.Through the analysis of some key algorithms in the process of emitter recognition,the parallel algorithm design is proposed.By comparing the running time of serial version and parallel version of each algorithm,the performance difference between them is analyzed.Firstly,the system composition and operation process of radar emitter recognition are introduced.At the same time,some key algorithms in the process of radar emitter recognition are introduced,including digital down conversion,digital channelization and pulse clustering sorting algorithm.The specific algorithms are digital down conversion algorithm based on low-pass filter,Digital Channelization algorithm based on polyphase filter and radar signal sorting algorithm based on DBSCAN clustering.Secondly,the GPU architecture and CUDA parallel programming are introduced.This thesis first introduces the hardware architecture of GPU,and then introduces a heterogeneous platform CUDA based on CPU+GPU,focusing on the programming model,memory structure,common mathematical library and optimization acceleration methods of CUDA.Thirdly,through the analysis of some key algorithms in the process of emitter identification,the feasibility of parallel design of the algorithm is obtained.The specific design scheme is further proposed,and the proposed design scheme is optimized.The main optimization methods are: using cu FFT library function,reasonable use of shared memory and manual optimization of kernel function.Finally,on the CUDA platform based on C language,each algorithm is programmed.GPU platform(NVIDIA geforce GTX 1660ti)is used to test the algorithm performance of the simulation data.Test results show that the designed algorithm can process the signal very quickly on the premise of ensuring the accuracy.Compared with the serial algorithm,the parallel algorithm effectively improves the speed of data processing and greatly improves the efficiency of the emitter recognition system.In the process of developing radar emitter recognition system,CUDA,a development tool based on heterogeneous platform,is used to improve the portability and scalability of the system.The results show that the running speed of the parallel algorithm designed in this thesis has been greatly improved,which has a certain engineering reference significance.