Research On Maneuvering Target Tracking Algorithm In Traffic Scene

Traffic target tracking is an important part of traffic information processing in the Intelligent Transportation System（ITS）.With the complexity of traffic environment,the continuous and stable tracking of traffic targets becomes a pressing problem to be solved.In this thesis,we focus on the mobile target tracking algorithm in traffic scenario,and focus on the near move-stop-move target tracking algorithm in traffic scenario,and the high mobile move-stop-move target tracking algorithm in traffic scenario,in order to achieve continuous and robust tracking of traffic targets.In this thesis,we address the difficulties of target miss detection and high mobility in traffic maneuvering target tracking,statistically analyze the motion behavior of traffic targets,propose the algorithms applicable to near move-stop-move target tracking and high maneuvering move-stop-move target tracking in traffic scenes,and verify the feasibility of the algorithms through simulation experiments and real measurement data.The main research content is divided into three parts.1、For the complex and variable behaviors of vehicle maneuvering targets in traffic scenes,analyze a variety of typical motion scenarios,propose a generalized target maneuvering motion model for traffic scenes,and analyze the set of applicable motion models in traffic scenes,and classify the maneuvering target tracking from shallow to deep into proximity move-stop-move target tracking and high maneuvering move-stopmove target tracking by modeling motion characteristics.2.To address the problem of mixed and lost motion measurements of approaching move-stop-move targets in traffic scenes,we deeply analyze the multi-target waiting light scenario and propose a space-domain correlation-based approaching go-stop-go target tracking algorithm,which is comparable to the nearest neighbor cheap joint probabilistic data association unscented Kalman filter（NNCJPDA-UKF）,the algorithm achieves the correct track matching,reduces tracking error,and improves target tracking success rate and tracking accuracy in the case of multiple targets lost in measurement.3.To address the problem of high maneuvering of high maneuvering walk-stop-walk target motion in traffic scenarios,we propose a high maneuvering walk-stop-walk target tracking algorithm based on road information cognition based on motion modeling and in-depth analysis of such scenarios as pull-over and intersection turnaround,which is more efficient than move-stop-move target tracking algorithm based on the Variable Structure Interacting Multiple Model（VS-IMM）.Compared with the move-stop-move target tracking algorithm based on VS-IMM,this algorithm achieves continuous tracking of multiple targets with multiple motion states,achieves correct track matching in the case of multiple target measurement loss,reduces tracking errors,improves the success rate and tracking accuracy of target tracking,and verifies the feasibility of the algorithm through real measurement data.The proposed series of algorithms have been verified by simulation experiments,with certain performance improvement and excellent performance in real measurement scenarios,which have certain engineering practice value.