Research On Target Tracking Technique For 77G Vehicle-Based Radar

As an important sensor in ADAS system,vehicle-based radar plays a vital role in driverless cars.In addition,target tracking is a major technical problem in the field of radar information processing to achieve various functions of the ADAS system.77 GHz millimeter-wave radar has higher detection accuracy and lower power consumption than 24 GHz millimeter-wave radar.While improving working performance,it also places higher requirements on 77 GHz millimeter-wave radar data processing.This thesis takes 77 GHz millimeter-wave vehicle-based radar as the foundation and makes following research work around the algorithm implementation and simulation of target tracking:(1)The structure of 77 GHz millimeter-wave radar system based on AWR1642 is analyzed.The method of acquiring the position,speed and angle information of the target by the vehicle-based millimeter-wave radar is studied.(2)Aiming at the characteristics of 77 GHz millimeter-wave radar that can acquire target point cloud data,clustering research is performed on 77 GHz millimeter-wave radar point cloud data to prepare for subsequent target tracking.It mainly includes the research and experiment of K-means algorithm,AGNES hierarchical clustering algorithm and DBSCAN algorithm based on point cloud location information.The thesis analyzes the shortcomings and deficiencies of the traditional clustering algorithm,and proposes a fast clustering algorithm of 77 GHz automotive millimeter wave radar based on Doppler frequency.Finally,the simulation experiments verify that the improved algorithm can improve the accuracy of point cloud clustering and reduce the time consumption caused by the increase of computing dimensions.(3)Aiming at the general road environment,this thesis studies and analyzes the vehicle-based radar target tracking technology based on data association.The track initiation algorithm is discussed and compared,including the intuitive method,logical method,Hough transform method and modified Hough transform method.Then the Kalman Filter(KF),Extended Kalman Filter(EKF)and Unscented Kalman Filter(UKF)are studied and simulated.Finally,the Near Neighbor Data Association(NNDA)algorithm,Probabilistic Data Association(PDA)algorithm and Joint Probabilistic Data Association(JPDA)in the track association algorithm are studied and simulated.(4)Aiming at the complex road conditions and shortcomings of traditional target tracking,the multi-target tracking technology based on Random Finite Set(RFS)is studied.The thesis focuses on the concepts of set integrals and set derivatives under the framework of RFS theory.By constructing a multi-objective state and observation model,the Probability Hypothesis Density(PHD)filter based on RFS theory is studied.According to the Gaussian mixture model(GMM),the filtering algorithms of the Gaussian Mixture Probability Hypothesis Density(GM-PHD)filtering algorithm,the Extended Kalman Gaussian Mixture Probability Hypothesis Density(EK-GMPHD)filtering algorithm and the Unscented Kalman Gaussian Mixture Probability Hypothesis Density(UK-GMPHD)are constructed under linear and nonlinear conditions.The simulation results show that the above algorithm has good multi-target tracking ability in complex road conditions.