Research On Distributed Multi-sensor Fast Fusion Algorithm Based On CPHD

Distributed multi-target tracking technology(DMTT)has attracted wide attention from academia and industry because of its low communication cost and strong system survivability.In recent decades,DMTT has achieved rapid development and continuous improvement,and there are still several problems and challenges to be solved.In this paper,the multi-target Bayesian filter based on random finite set theory is the subject,and the distributed fusion algorithm of Cardinalized Probability Hypothesis Density filter,the fast implementation method of fusion algorithm and the network consensus algorithm are mainly studied,which provides the foundation and technical support for the application and implementation of DMTT technology.The main work is as follows:1.Aiming at the problem of multi-target tracking,the multi-target Bayesian filter based on random finite set theory is studied.The Cardinalized Probability Hypothesis Density(CPHD)filter is used to model the problem of multi-target tracking,and the distributed CPHD fusion algorithm based on Generalized Covariance Intersection(GCI)criterion(GCI-CPHD)is studied,which provides a foundation for further research of distributed fusion algorithm.2.Aiming at the computational complexity of GCI-CPHD multi-target fusion algorithm,a parallel GCI-CPHD fusion algorithm based on Union-Find fast joint grouping(G-CPHD)is proposed,which can greatly reduce the computational load of fusion tracking by abandoning the fusion components with low weights to approximate the fusion posterior distribution.3.To solve the problem of fast computation of global consensus of posterior distribution of multi-sensor fusion in distributed network,a parallel CPHD distributed fusion consensus algorithm(C-G-CPHD)is proposed.The algorithm adopts the strategy of sequential information interaction between adjacent nodes and the weighted Kullback-Leibler average calculation,which makes all the nodes of sensor network have the ability of information sharing and global estimation.The feasibility and effectiveness of the above-mentioned algorithm and its numerical calculation method are verified by simulation experiments.