Multi-Radar Based Target Tracking Method Of Intelligent Vehicle

Data association and motion state estimation are two difficulties in the existing target tracking methods of intelligent vehicle.On the one hand,the data association process only considers the motion parameters such as the target position and the velocity,it is easy to have a wrong association in a complex environment.On the other hand,the target is modeled as a point in traditional state estimation method,which is not enough to fully describe the target motion process.Therefore,the research on target tracking method with good stability and ability to meet the needs of intelligent vehicle target tracking task in the actual environment has theoretical significance and practical value.Aiming at the data association problem,and to improve the accuracy of moving target recognition.Firstly,a local grid map of intelligent vehicle is built in this thesis,and the grid map is processed by closed operation to ensure more accurate grid clustering and contour extraction.Secondly,using LiDAR raw data to detect the road edge and define the road passable area.Thirdly,distinguishing between moving targets and stationary targets based on target shape feature,motion feature and road passable area,so as to reduce the false detection rate of the moving target.Finally,based on the optimal allocation theory,the target feature information is introduced as an auxiliary condition to improve the data association result and the real-time performance of the algorithm.In order to improve the target filter model,the change of target feature information during movement is considered in this thesis,and a hybrid tracking method of point target and extended target is proposed.Firstly,classify moving targets into point and extended target based on feature size and distance of target.Then,the corresponding state space model is established to describe each target.Thirdly,building a point target and extended target hybrid tracking filter based on the Kalman filter framework,which keeps tracking point targets and extended targets at the same time,and achieves tracking model switching adaptively for the same target at different times.Firstly,Simulation scene experiment is designed to verify the validity and stability of the proposed method.Then,a target tracking system software is designed and implemented,which encapsulates all function modules and only user interface is provided.Finally,the system is used in real vehicular experiment,experimental results shows that its performance is good in time and stability,it also proved the stability and effectiveness of the proposed method.