Research On Target Tracking And Classification Algorithm Of Millimeter Wave Traffic Radar Targets Based On LMB Filter

Accurate road information perception is one of the basic core technologies of intelligent transportation system.As an important module,traffic target tracking and classification technology,whose task is to obtain real-time target measurement information through various sensors such as radar and camera distributed at urban intersections and road sections,which can be used as the input of tracker and classifier,respectively,in order to obtain kinematic information such as vehicle orientation,speed,etc.and distinguish the types of targets.To sum up,the tracking and classification technology of traffic targets has important engineering application value and practical research significance.Based on LMB filter and the theoretical framework of joint tracking and classification,thesis studies the target tracking and classification algorithm of millimeterwave traffic radar.The main work is as follows:1.Aiming at the problem of traffic target tracking,the random finite set theory is briefly introduced,including several commonly used random finite sets and labeled random finite sets,with emphasis on the multi-objective filtering framework based on random sets.Under the millimeter-wave traffic radar system,the motion state model,the motion transfer model and the millimeter-wave traffic radar target measurement model are constructed,and the prediction and update process of LMB filter based on these models is deduced.The tracking performance in complex traffic environment is better than that of traditional tracking algorithms.2.Aiming at the problem of vehicle-road interaction in traffic environment,a millimeter-wave traffic radar target tracking algorithm based on vehicle-road interaction model is proposed.The Gaussian mixture model is used to approximate the process noise of traffic targets,and the vehicle-road interaction behavior is simulated by adaptively adjusting the weight of Gaussian components of noise,which can effectively deal with the vehicle-road interaction problem and has higher tracking accuracy than the traditional LMB filter.3.Aiming at the classification of targets in traffic environment,the characteristics of traffic targets under millimeter wave traffic radar system are analyzed,and the extended dimension classification attribute of the combination of target size features and kinematics information features is proposed,which can realize the effective classification of traffic targets.An adaptive connected domain clustering algorithm suitable for traffic targets is proposed,which can realize the effective clustering of multi-measurement points of adjacent extended traffic targets,and is more effective than traditional target aggregation algorithm and classical connected domain clustering algorithm.4.Aiming at the coupling problem of traffic target tracking and classification tasks,under the framework of joint tracking and classification theory,a joint tracking and classification algorithm of millimeter wave traffic radar based on LMB filter is proposed.In the prediction stage,different filter parameters are used for different types of targets by using type-related filter parameter sets to match the real motion of the targets more closely;In the update stage,the measurement likelihood equation of target and clutter is extended by using the size characteristics of traffic target,and the correlation accuracy between target and measurement is improved.The above work has been verified by simulation experiments and measured data,which can realize the function of traffic target tracking and classification and solve the problem of traffic target tracking and classification.