Research On Key Technologies Of Multi-target Tracking Of Millimeter-wave Radar

Millimeter wave radar is widely used in the fields of intelligent transportation and automated driving.It can be used for the statistics of traffic flow,the dangerous warning of the blind area at the intersection,and the advanced driving assistance system(ADAS)of automobiles.As an important part of millimeter wave radar,multi-target tracking technology can identify and track illegal and speeding vehicles on the road,and strengthen the monitoring of road traffic.It can also provide accurate information for ADAS to ensure the safety of autonomous driving.Therefore,the research on multi-target tracking technology has important practical significance.This paper is engaged in further research on the multi-target tracking technology of millimeter wave radar.The specific work content is as follows:First,the issue of radar measurement data preprocessing and track initiation are studied,and an outlier elimination method based on improved Kalman filtering is proposed,which uses the ratio of the expected value and the predicted value of the measured innovation covariance matrix as a correction factor to dynamically correct the measurement,reduce the influence of outliers and adjust the innovation.The simulation shows the method can effectively suppress the influence of outliers and reduce the filtering error.Then,based on the density-based spatial clustering algorithm(DBSCAN),a method for cumulative clustering of radar data of N frames is designed.And on this basis,a path initiation method based on logic method and finite state automata is designed.This method can effectively converge the measure,initializing track and reducing the number of point to be processed by 1/3.Secondly,the issue of data association between the track and the point is studied.An improved MHT data association algorithm is proposed based on the multiple hypothesis tracking(MHT)algorithm.This method reduces the size of the association matrix by simplifying the association matrix,and uses the Jonker-Volgenant Algorithm for Linear Assignment Problem(LAPJV)method to improve the speed of solving the optimal solution of the association matrix.The simulation shows that the operation speed of the improved MHT association matrix is about one tenth of that of the classic MHT.Then,based on the previous content,a multi-target tracking algorithm ST-IMM-MHT based on a single track-initiation and improved MHT is proposed.This method separates the track-initiation from MHT,uses the method for track-initiation designed in this paper to process separately,and uses the improved MHT and IMM to combine data association and filtering.The simulation shows that under high clutter density,compared with the classic IMM-MHT tracking method.ST-IMM-MHT can reduce computing time by about 80%.Finally,the ST-IMM-MHT is verified through experimental data in different scenarios,and the results indicate that it can effectively track multiple targets within the radar detection range.