Research On Signal And Data Processing Technology Of Millimeter Wave Radar Based On GPU

In recent years,with the continuous development of autonomous driving technology,high-performance on-board sensors have become an indispensable research direction.Among them,vehicle-mounted millimeter wave radar,with its working ability in any weather or time,even in rain,snow,fog and other harsh environment,still maintain stable performance,and has the advantages of small size,high integration,low cost,has become one of the core sensors of automotive automatic driving system.Therefore,based on vehicle-mounted millimeter wave radar,it is of great practical significance to study the high-precision and high-efficiency detection technology for targets under complex road conditions.CUDA parallel computing program based on GPU brings a new optimization direction for radar signal processing algorithm.Based on Linear Frequency Modulated Continuous Wave,this thesis analyzes target detection methods such as speed measurement,distance measurement and Angle measurement,as well as its efficiency optimization scheme through GPU parallel computing.The algorithm of tracking the detection results is studied,which provides effective technical support for the practical application of millimeter wave radar.The main work of this thesis is as follows:1.This paper firstly introduces the signal model of LFMCW,and the principle of distance measurement,speed measurement and Angle measurement based on the signal.Then,the signal processing algorithm of MIMO radar based on LFMCW is studied,including the processing model and steps of range dimension FFT,velocity dimension FFT,Angle dimension FFT and two-dimensional constant false alarm detection（2DCFAR）,as well as the special processing steps of MIMO radar signal processing,such as channel correction and Doppler phase compensation.2.The theory and method of target tracking after signal detection are studied,including target state prediction,correlation and tracking,as well as feature extraction method based on target radar scattering interface（RCS）,so as to realize the classification and recognition of pedestrian,bicycle,car and other targets in the driving road,as well as their respective track tracking.3.Due to the real-time requirements of millimeter wave radar,each module of radar signal processing algorithm is analyzed,and the GPU-based parallel calculation method is implemented respectively,including channel correction,range dimension FFT,velocity dimension FFT,Doppler compensation,Angle dimension FFT,2D-CFAR,peak detection and other modules.The accuracy and acceleration effect are evaluated,and it is proved that the proposed method can meet the real-time requirements of actual millimeter wave radar system.