Research Of Pulse Doppler Radar Signal Processing Based On Parallel Technology

With the wide application and the increasing complexity of pulse Doppler radar,the signal processing trends to be digital,software,intelligent and multi-processing system integration,which makes the radar processing system facing the problem of large amount of data and high real-time requirement.Research on parallel technology of radar signal processing can make full use of the parallelism of signal processing algorithm itself to speed up the signal processing.This thesis introduces the parallel technology of pulse Doppler radar signal processing and the concrete content is as follows:Firstly,the thesis introduces the overall framework of pulse Doppler radar signal processing system,which includes pulse compression module,MTI module,MTD module and CFAR detection module.Among them,the realization of CFAR detection module is mainly studied,including one-dimensional CFAR algorithm,two-dimensional CFAR algorithm,clutter point detection algorithm and clutter surface detection algorithm.The thesis introduces the simulation implementation and analysis of every module.Secondly,the thesis introduces several parallel platforms in signal processing,including FPGA,DSP and GPU.Comparing the advantages and disadvantages of their internal modules,parallel implementation,software programming model and radar signal processing algorithm,the thesis determines to choose CPU+GPU as the hardware platform.What's more,the thesis introduces more details of hardware and software model of the chosen platform.Thirdly,the thesis researches the key algorithms of pulse Doppler radar signalprocessing system based on CPU+GPU heterogeneous platform.Including pulse compression,one-dimensional CA-CFAR,two-dimensional CA-CFAR and clutter map dot detection technology.The thesis analyzes the parallelism of each algorithm and programs each algorithm with CUDA programming model.The thesis carries out seveal experiments to analyse the effects of some factos on the acceleration ratio of the program,including the partition of threadsPerBlock,the improvement of the algorithm,the memory optimization and expansion of the data.The results of the experiment confirmes that the parallel processing method based on GPU can achieve a maximum speed of 106.72 times compared with CPU.