Research And Implementation Of Tracking Algorithms For Traffic Millimeter-Wave Radar

Millimeter-wave radar refers to radars that operate in the frequency between 30GHz to 300GHz and have a wavelength between 1mm to 1cm.It has high range accuracy and speed accuracy,strong anti-interference ability,and can work all day.So it is widely used in the field of transportation.Multi-target tracking is an important part of the signal processing of millimeter-wave radar,which can make sure that in the multi-target scenario,the detection points of the vehicle can be associated with its own track as many as possible.In traffic scenario,vehicle angle measurement information is very significant.With the lack of cameras,the measurement angle of the vehicle directly determines which lane the vehicle is driving in.Then we shall determine whether the vehicle is retrograde,whether there are dangerous vehicles nearby,and get other information such as traffic statistics.The research on tracking algorithms and the work of this paper are as follows:(1)The process of the tracking system and its various tracking algorithms are researched,including point agglomeration,the selection of tracking gates,speed ambiguity resolving and track state management.Experiments are carried out on most of the algorithms to show their effectiveness.Data association algorithms,including the nearest neighbor data association algorithm and the probabilistic data association algorithm,are focused in the paper.Their advantages and disadvantages are analized by experiments with actual data.(2)The motion models of the target and Kalman filtering algorithm used after targets’ association are introduced.Various derivative algorithms of the Kalman filtering algorithm are introduced,including the linear Kalman filtering algorithm used in linear models,extended Kalman filtering and unscented Kalman filtering used in nonlinear models.The advantages and disadvantages of each derived filtering algorithm are analized by experiments with simulated data or actual data.The adaptive part of the Kalman filter is introduced,of which the effect is demonstrated through experiments with simulated data.The effects of the initial noise matrixs on the filtering effect are importantly researched by formula derivation and actual data experiments.(3)Vehicles in traffic scenario can travel up to several hundred meters and as the distance increases,the signal-to-noise ratio of the radar decreases.Millimeter-wave radar will experience large jitter in angle measurement in the distance.Therefore,the multi-target tracking algorithms in the traffic scenario are focused in this paper,especially the angle filtering algorithm.In order to solve the problems existing in the project,a tracking system is set up based on theoretical knowledge,an appropriate motion model and a filtering type are selected for targets in traffic scenario and appropriate tracking algorithms are added to the system.The filtering method with distance filtering and angle filtering is improved from the model in which the horizontal and vertical portions of the Cartesian coordinate system are separately filtered.The improved method makes the filtered trajectory more consistent with the true trajectory of the vehicle target,reduces the angle jitter,improves the correlation effect of the trajectory and enhances the reliability of lane identification using angles when the lane lines are known.The correlation between the mirrored false alarm data and the real target data is analyzied by collecting actual data,promoting the tracking algorithm adjusted on Matlab to filter out most of the mirrored false alarms and reduce their interference in road condition judgment.