Dynamic Target Motion Information Extraction Technology Based On Laser Point Cloud

With the perpetual development of modern industry and artificial intelligence technology,people are increasingly dependent on intelligent products.The innovation of modern technology has made computer vision widely concerned by people.In the field of computer vision,the identification and tracking technology of dynamic targets has become a hot issue for many researchers.Target tracking technology plays an important role in the field of intelligent monitoring,in addition,it has very practical value in assisted driving and automatic driving.So far,researchers have developed many results on target tracking,and target tracking technology has made breakthrough progress.The relationship between target tracking technology and image processing technology is inseparable.Compared with 3D point cloud images,the application of target tracking technology in the field of video images is relatively mature.However,the video image lacks the depth information of the target object,so occlusion between the target objects may occur frequently,and that causes certain difficulties in target detection and tracking.Aiming at the shortcomings of traditional video images in the field of target tracking,this paper proposes a 3D target tracking method based on LiDAR to collect scene information.Firstly,we choose the particle filter algorithm that can deal with nonlinear non-Gaussian systems in the choice of tracking algorithm.Based on the particle filter algorithm,this paper proposes an improved particle filter algorithm--Auxiliary regularized particle filter(ARPF)algorithm.First of all,the algorithm introduces an importance density function for secondary weighting,and then improves the selection of resampling,which is by the discrete sampling from the posterior probability density to the approximate continuous distribution.Using MATLAB to do a simulation test on the ARPF algorithm proposed in this paper,the effectiveness of the algorithm is proved by testing.To realize real time target tracking based on laser point cloud,PCL(Point Cloud Library)is introduced specially for pretreating and dealing with point cloud.Then process framework for point cloud target tracking is proposed based on PCL,which provides a theoretical basis for point cloud target tracking in the real scene.In the real time target tracking experiment,the laser radar VLP-16 is used as the acquisition device.The PCL is used to track the dynamic target of the point cloud dataset under the ROS(Robot Operating System)framework,and finally the motion information of the target object is extracted by calculating the velocity and acceleration of the target object.The experimental result shows that the improved particle filter algorithm proposed in this paper can not only meet thereal-time performance of tracking,but also has better robustness in the case of interference among target objects and change in motion speed.