Study On Target Detection And Recognition Technology In Vehicle-Mounted Millimeter Wave Radar

In the application of driverless vehicle,the joint data fusion signal processing of a variety of different sensors is often used.The millimeter wave radar is a sensor that uses millimeter wave band for active sensing,and it is also one of the sensors with strong anti-jamming ability and suitable cost in the driverless equipment.Compared with optical cameras,lidar and other sensors,millimeter wave radar is not limited by light,meteorological environment and other conditions.It has strong ability to penetrate dust,haze and smoke,and has the characteristics of all-weather and all-time operation.It can track and identify interested targets,and has high data accuracy and accurate pitch angle,The mode transformation can be carried out according to the demand to achieve the short,medium and long-distance detection and the demand change of different resolutions.In the future application of driverless technology,millimeter wave radar will become an indispensable part.At present,millimeter wave radar detection has been widely used in driverless,traffic and road safety,smart home,security detection,military reconnaissance and other fields.This thesis focuses on the research of millimeter wave radar in frequency modulated continuous wave mode,and improves the algorithm in the existing methods:1.To solve the problem of low accuracy of angle measurement in 2D-CFAR detection framework,the traditional phase angle measurement method is replaced by the high-resolution direction measurement method based on subspace structure,and the target information matching mechanism is used.In the case of obtaining the target condensation results after 2DCFAR detection and the target physical information results calculated by the unit index,the phase angle measurement method is used,feedback to the super-resolution algorithm and matching mechanism,through the joint matching of space-time and time-domain information to improve the accuracy.2.In order to improve the traditional method and process of measuring distance,speed and angle in FMCW mode,sparse fast fourier transform is added to replace the traditional transform method.By combining the information of fast time dimension and slow time dimension,2D-CFAR and SFFT are combined to reduce the amount of calculation and noise information through spectrum compression,improve the detection performance.3.In the target recognition of millimeter wave radar,the motion model of different targets is established,and the micro doppler features of targets are extracted by combining multi frame radar echo data.Through time-frequency analysis of multi frame data of tracking targets,multi frame joint analysis of radar echo data in FMCW mode is carried out by generating gray level co-occurrence matrix,the time-frequency variation features of different objects due to different motion patterns are extracted,and the features are put into the trained neural network to achieve the purpose of target recognition.Based on the traditional algorithm,this paper combines the super-resolution algorithm with the phase comparison method in the framework of 2D-CFAR through the target information matching mechanism to improve the angle measurement accuracy;the 2D-CFAR based on sparse fourier transform can reduce the influence of noise on target detection and reduce the amount of calculation;on the basis of the above,the micro doppler information of the target is analyzed,and the target classification is achieved by combining the micro doppler time-frequency texture structure with long short term memory framework.