| Digital array radar plays an important role in array radar,which can form several beams at the same time to realize omnidirectional scanning.As a form of digital array radar,cylindricalarray radar has the characteristics of omnidirectional multi-beam and large data rate,which places high demands on signal processing.Conventional radar signal processing platform is mainly based on FPGA and DSP,with high development cost and low versatility.Due to the large amounts of parallel computations in signal processing,this paper has carried out the research on cylindrical-array radar signal processing related algorithms based on GPU parallel parallel computing platform.The CUDA software programming technology has been used to realize the transplantation of the algorithm from the FPGA+DSP platform to the GPU platform.The main work and contributions of this thesis are divided into three aspects.1.Task analysis of cylindrical-array radar,design and verification of signal processing.This radar system is aimed at the detection and tracking of low-altitude and low-speed small targets.In order to take the ability of detecting both far and close targets into account,a widenarrow pulse composite signal is used as the transmission signal.Combined with the index of radar system,the signal processing flow based on 36 channels is designed.Each 12 channels is a group,the left and right beams are formed successively for pulse compression,MTD and CFAR processing.The signal processing algorithm was simulated in Matlab environment.2.Scheme design and signal processing algorithm implementation based on GPU platform.After a thorough study of the GPU architecture and CUDA programming model,the complexity and parallelism of signal processing algorithms are analyzed,and a general scheme of signal processing based on the GPU platform is proposed and demonstrated.The system adopts cooperative and heterogeneous working mode,CPU is responsible for task scheduling,and GPU is responsible for parallel data operating.The design of signal processing algorithm is divided into modules,which adopts the method of serial between modules and parallel in module.The operation types of each algorithm module and the scale of data to be processed are analyzed,and the kernel-level parallelism of the signal processing algorithm is implemented.3.Software implementation and algorithm testing of cylindrical-array radar based on GPU.This paper introduces the overall architecture of cylindrical-array radar system,and gives the realization method of cylindrical-array radar data transmission,signal processing and display control terminal.The 36-channel echoes generated by the echo simulator are transmitted to the server through the PCIe 3.0 bus,then GPU parallel algorithms such as FIR filtering,DDC,DBF,pulse compression,MTD and CFAR are tested.Compared with Matlab results,the error of test results meets the design requirements,and the reliability of the algorithm is verified.The data processing time in a CPI meets the real-time requirements of the system,and the feasibility of the algorithm is verified.This paper finally completed the implementation of cylindrical-array radar signal processing algorithm based on GPU platform,and the test results were given,which solved the problems of high development cost,long development cycle and low versatility.The experimental results show that the signal processor based on GPU platform has high reliability and strong real-time performance,which can meet design requirements,and provides a reference for the continuous improvement. |
PDF Full Text Request