Implementation Of Radar Signal Processing Algorithms Based On GPGPU

In recent years,with the emergence of various types and functions of drones,the use of drones has become the primary problem of security in places such as airports and large open-air venues.Low-altitude slow small target detection represented by drones has also become one of the important problems in radar detection.Although airborne early warning radars and ball-borne radars have the advantage of detecting low-altitude targets,they are complicated,difficult to implement and costly.Traditional digital signal processing platforms such as FPGA+DSP has great difficulty in hardware and software development.The development cycle is long,not easy to maintain and not easy to port.Therefore,this thesis designs a radar system based on the general-purpose parallel computing platform GPGPU,which is based on the linear frequency modulated continuous sawtooth wave as the transmitting waveform for the low-altitude slow small targets.The research method used in this thesis is to first explain the principle of each signal processing algorithm and the target information extraction process through MATLAB simulation.Then,the feasibility of signal processing of linear frequency modulated continuous wave radar on GPGPU is studied.The MATLAB results are compared to prove the correctness of the GPGPU implementation.Finally,the same radar signal processing algorithm is implemented on the CPU by means of the MKL library function,and the implementation time of the two platforms is measured separately.The time-comparison illustrates the acceleration effect of the parallel implementation of the GPGPU platform.The thesis first introduces the basic composition and principle of the linear frequency modulated continuous wave radar system and the principle and effect comparison of the digital signal processing algorithm.The formula deduces how the chirped continuous wave radar uses the beat signal of the local oscillator signal and the echo signal to obtain the velocity and distance information of the targets,and how to measure the target angle by the beam scanning method.Secondly,the unified parallel computing architecture CUDA is briefly described.CUDA is a framework provided by NVIDIA for the mixed programming of CPU+GPU heterogeneous platforms.It integrates some interface functions and general function libraries to facilitate rapid program development.Finally,the parallel implementation process of the chirped continuous wave radar signal processing algorithm(pulse compression,digital beamforming,clutter suppression and moving target detection,constant false alarm detection,and dot condensation)on the CPU+GPU heterogeneous platform is completed.It can be seen from the target track detection result of the display and control terminal that the radar system has the ability to detect low-altitude slow small targets in real time.Through the same algorithm,the comparison of the operation time of different hardware platforms proves that the CPU+GPU heterogeneous platform has obvious acceleration effect,which has great significance for the real-time performance of the radar signal processing system.