Research On Extended Object Tracking Technology Of 77G Millimeter Wave Radar

Compared with other sensors,77 GHz millimeter wave radar has the advantages of strong penetrating power,less susceptible to environmental influences,and ensuring privacy.But compared to fields such as video tracking and other fields,most of the existing researches on the extended target tracking algorithm for millimeter wave radar are based on simulation,which lack of a system-level framework that can provide sensor data interface,target tracking algorithm implementation,and metric implementation.This not only makes the theoretical research unable to be quickly applied to actual projects,but also the performance evalutioin of the tracking algorithm is hard to apply the measured data.Based on the above problems,in this thesis,an extended target tracking framework with high cohesion and low coupling is constructed based on the object-oriented idea.It provides a unified measured data interface while realizing the target tracking algorithm,and without changing the frame structure,new target tracking algorithms and sensor data interfaces can be flexibly introduced.While realizing code reusability and inheritance,it also lays the foundation for multi-sensor data fusion.In addition,according to the highresolution characteristics of 77 GHz radar,an extended target tracking system based on group centroid tracking is also studied.The system can eliminate static targets and realize single target and multi-target tracking.In this thesis,the following works are carried out in the aspects of algorithm research,experimental platform design and construction,data preprocessing,framework design,simulation verification and experimental verification.1.Firstly,the signal processing and data preprocessing methods of 77 GHz millimeter wave radar are studied,and then multi-threaded point cloud data acquisition system is built up.Finally,in order to cluster the extended target into point targets,two point cloud data clustering methods are analyzed and a DBSCAN clustering algorithm based on multi-frame point cloud data is proposed.2.In order to solve the problem that the existing target tracking methods lack the measured data interface,in this paper,we cluster the extended target into point target,and concentrate the target tracking problem based on millimeter wave radar into a unified tracking framework.Thus,a target tracking framework with object-oriented characteristics is realized.The framework defines different system models,data association algorithms,metrics and trajectory management algorithms in the form of classes.In addition,it provides different sensor data interfaces to facilitate the application of new tracking algorithms to vehicle-mounted millimeter-wave radars,and lay the foundation for the research of multi-sensor multi-target tracking algorithm.3.Aiming at the disadvantage that point target can not provide target range,and considering the similarity between group target and extended target,this paper studies an extended target tracking framework based on group centroid tracking.The uncertainty of group target motion and the dispersion degree of group are considered when updating the target state.Finally,the millimeter-wave radar IWR1642 is used to track single pedestrian and multi-pedestrian targets,which verifies the effectiveness of the algorithm.4.Finally,a measured data verification platform based on the 77 GHz millimeter wave radar IWR1642 and the 4D imaging radar IWR6843-ODS is built,and carried out the actual test verification and effect evaluation of the extended target tracking framework.The experimental results show that the framework achieves robust tracking of single and multiple pedestrian targets in indoor scenes.