| In recent years,the road traffic environment has become increasingly complex.At the same time,the management requirements for intelligent transportation systems in China are also more perfect.Their functions are required to be increasingly influential in terms of flexibility,coverage,detection results and real-time.It is clear from the research that the detection methods of big data,video and radar can realize the judicious management requirements for the transportation system.Millimeter wave radar has a little volume,convenient installation mode and strong adaptability to rain,snow and fog,so it also has extremely important applications in the field of intelligent transportation.Its most representative value includes multi lane detection,which can automatically identify the coming and going direction of vehicles according to the distance and speed of vehicles in different detection lanes;provide real-time statistics of lane occupancy and traffic flow,classify the monitored vehicles;track vehicle trajectories and deal with abnormal traffic phenomena such as illegal lane change in time.Target tracking is the most important part of radar data processing.With the increasing traffic flow on daily traffic roads,the stability and real-time performance of tracking algorithms have become a problem that has to be broken through.The aforementioned single target tracking can no longer meet the requirements of the current situation.So it is imperative to develop a stable tracking algorithm that can be adapted to multiple targets.In this thesis,a target tracking platform is constructed for the application of millimeter wave radar in transportation.The filtering and data association algorithm is proposed as the framework of the tracking algorithm in this paper,and the algorithm is simulated,compared and optimized.Finally,real data obtained by signal processing is sent to the data processor for practical engineering verification,and a multi-target tracking system with excellent performance is developed,which is expected to promote the further development of China’s transportation industry.The main work of this thesis is as following:(1)A systematic analysis of the research background of Millimeter wave radar target tracking,the current state of development at home and abroad and the problems that exist in the current research process is presented,an improved algorithm applicable to engineering is proposed based on the existing problems.(2)Aiming at the problem of system modeling in target tracking,this paper first focuses on three motion models with different states,namely: constant velocity(CV),constant acceleration(CA)and constant turning(CT).Then the Kalman filter which is adapted to the linear motion model is simulated in this paper.Considering the nonlinearity of most target motion models in the actual road conditions,the extended algorithm of the Kalman filter is proposed,which not only improves the filtering accuracy but also makes the simulation results more practical.Aiming at the matching problem of multi-point traces and multi-wave gates that may occur in multi-target tracking,this paper proposes three data association algorithms,namely Nearest Neighbor Data Association(NNDA)and Probabilistic Data Association(Probabilistic Data Association).Association,PDA)and Joint Probability Data Association(JPDA).Through the Root Mean Square Error(RMSE)value obtained by simulation,it can be seen that the joint probabilistic data association algorithm has a better tracking effect for multi-target.(3)In the case of an increase in the number of targets,the computational effort of JPDA algorithms can skyrocket,hindering practical applications in engineering.The main idea of the improved algorithm is to continuously decay the probability value of the common region to carry out the state update of the target,avoiding the computational effort caused by splitting the matrix and reducing the computation time,which is more conducive to engineering practice.(4)For the study of multi-target tracking algorithms under complex roads,a target tracking technique with the idea of mathematical sets is proposed.The advantage of this algorithm is the fact that it discards the traditional data association operation of multiple tracks and point traces,so the computational efficiency is significantly improved.Then Probabilistic Hypothesis Density Filter(PHD)based on random set theory is investigated,and the good performance of the above algorithm in multi-target tracking is verified by simulation using Optimal Sub-patten Assignment(OSPA)as the evaluation index.(5)The working mechanism of Millimeter wave radar is introduced.To better study the characteristic information of distance,speed and angle required for target tracking,the signal processing process of the radar is analyzed in detail,and the real data collected by ADC is verified on Matlab,which makes a theoretical basis for subsequent target tracking.(6)The traditional track initiation algorithm is studied.On the basis of the traditional track initiation algorithm,an improved tracking initiation algorithm based on the logic method is proposed for targets with strong maneuverability,and the good effect of the track initiation is demonstrated through simulation.Focusing on the track management problem,the customary track quality management method is optimized and the quality of the target track is evaluated by the scoring method.(7)A system implementation of the target tracking algorithm is presented,including frontend introduction and platform design.Finally,based on the ZYNQ platform,the system is tested on a large number of roads.The test scenario is to set up a high radar at the traffic lights of an eight-lane city road,and the algorithm is analyzed by capturing the target motion information in real-time.From the results,it can be seen that the radar can achieve multitarget tracking and present stability and real-time,with an accuracy rate of up to 95%,which verifies the feasibility and validity of the relevant theory in this thesis. |
PDF Full Text Request