Real-time Tracking Technology Of Dynamic Target Based On Laser Point Cloud

In order to reduce traffic pressure,improve the traffic environment,and improve traffic safety,people have begun a lot of research on unmanned driving.Among them,in the tracking of unmanned driving targets,in order to judge the movement of the target in time and accurately,usually choose the camera,Millimeter-wave radar,lidar and other equipment collect information.Among them,lidar is regarded as one of the most important information collection equipment because of its high measurement accuracy and strong environmental adaptability.In this paper,the target information is collected by lidar,and related research such as target tracking is carried out on the basis of its scanning data.The main research contents are as follows:In the target detection process,the Voxel Grid filter is first used to down-sampling the point cloud data to ensure that the point cloud density is reduced when the point cloud characteristics are unchanged,and then the ground point cloud is segmented and removed using a plane fitting algorithm and statistical filtering is used to remove noise points from the point cloud,in order to cluster the processed point cloud more accurately,an improved clustering algorithm is proposed,that is,based on the Euclidean clustering algorithm,different regions are set to set different clusters.The method of threshold value of class radius,and the reflection intensity is introduced to judge when the people and cars are easy to be clustered into one class.In the target tracking process,in order to estimate the target state better,this paper proposes an improved particle filter algorithm based on the particle filter algorithm,combined with the regularization idea and lossless Kalman filtering,that is,the selection of the importance density function use lossless Kalman filtering to update the state,use regularization steps to optimize particle resampling during particle resampling,and use the least skew simplex sampling strategy to select Sigma points in lossless Kalman filtering,so as to ensure the estimation of performance and reduce the calculation time.In terms of data association,in order to achieve real-time target tracking,this paper selects the measurement by setting thresholds on the position and velocity in the joint probabilistic data association,and attenuates the association probability between the measurement and each target.