Multi-Information Fusion Target Tracking Based On Kalman And Particle Filter

In systems such as satellite navigation and positioning,radar tracking and guidance,automatic control systems,vehicle traffic planning,medical target recognition,robot mobile positioning,temperature prediction control,etc.,it is necessary to estimate the state parameters of the desired object by using the corresponding filtering techniques on the observed data and system state relations with known parameters,and this research hotspot is called target tracking technology.Target tracking is the process of state modeling,filtering estimation and continuous tracking of moving targets by using various observation and computational methods.There is a wide range of filtering techniques and tracking systems for different target systems,but better development is needed for overall tracking accuracy,real-time and robustness.In this thesis,a hybrid algorithm of traceless Kalman filter and particle filter is proposed for trajectory tracking,and a multiobserver information fusion technique is proposed to cope with the multi-target tracking problem.(1)To solve the problem of expanding the application range and improving the tracking efficiency of filtering algorithm target estimation.At the theoretical level,based on the traditional linear Kalman algorithm(KF),extended Kalman algorithm(EKF),traceless Kalman algorithm(UKF),and particle filter(PF)algorithm filtering,the concept of PF filtering as the front-end input processing,followed by UKF structure as the filtering framework to get the output,constitutes the hybrid algorithm of traceless Kalman filtering and particle filtering assembled in this thesis.The application of the hybrid filtering based on the traceless Kalman algorithm and the particle algorithm in different mathematical model scenarios under single observation and single target is investigated.The algorithms of EKF,UKF,PF and this thesis are applied to pure orientation target tracking experiments,and the results show that the improved filtering in this thesis has good advantages in tracking performance.Then the pure distance target tracking and temperature real-time prediction systems,using the UKF algorithm with better filtering performance as well as the PF algorithm are compared respectively.Judging from the two dimensions of tracking accuracy and real-time,the hybrid algorithm improves the tracking accuracy by 53.6%,36.8%,and13.9% compared to the EKF,UKF,and PF algorithms,respectively,within satisfying the time delay,which reflects the superiority in the target tracking problem.(2)To solve the problem of accurate information processing and efficient tracking for multi-target tracking application scenarios,a hybrid filtering multi-target information fusion tracking system is proposed.To study the experiment of multi-target tracking under multiple observatories based on UKF-PF hybrid filtering,multiple observatories collect unknown target arrays in known areas,use the nearest neighbor method to correlate and classify the data,constitute the corresponding target observation information,filter the real-time data with UKF-PF algorithm to get more optimized estimation information,and take the advantage of multi-information fusion to adopt weighted fusion for multiple target tracking processing to get the fused trajectory routes of different objects.The multi-information fusion system in this thesis can implement target tracking for multiple targets,and can simultaneously detect a single independent target and weight it to get a high-precision fusion trajectory.In case of damage or paralysis of a single observer or a large tracking error,the multiobserver information fusion improves the robustness of the system.The experimental results show that the multi-observatory has absolute accuracy advantage for single or multiple targets,and has certain reference value for engineering applications.