Group Target Tracking Fusion Technology Based On Generalized Label Multi-bernoulli Filter

Target tracking is an important issue in modern warfare.With the continuous development of military equipment and the increasing complexity of the electromagnetic environment,target tracking is facing ma ny problems.Dense multi-target tracking and group target tracking in complex scenarios have gradually become research hotspots in the field of target tracking.According to the difference of radar measurement accuracy in practical applica tions,this paper adopts the generalized label Bernoulli(GLMB)filter to study the tracking fusion technology of distinguishable group and indistinguishable group.The main research contents are as follow:(1)Resolvable group target tracking.For distinguishable group targets,it is not only necessary to estimate the motion state of all targets in the group,but group structure information is very important.In this case,it is considered to estimate the group structure and the motion state of each target in th e group.In this paper,the evolutionary network is used to model the distinguishable group structure.Simultaneously,the tag RFS is used to solve the target identity problem,and the GLMB filter is used t o track the distinguishable group.(2)Unresolvable group target tracking.Aiming at the indistinguishable group target,it is difficult to distinguish the measurement source within the group.In this case,it is considered to estimate the group centroid and its diffusion shape.the paper studies the adaptive star-convex random hypersurface(RHM)model to model the group diffusion shap e,solving the problem that the circular priori RHM is difficult to approximate the group shape quickly and accurately.Meanwhile,mean shift(MS)algorithm is used to divide the group target measurement to reduce the amount of calculation,and second-level group structure is established to improve the accuracy of the group number estimation when the group distance is relatively close.Finally,the GLMB filter is used to track the indistinguishable group.(3)Research on group target distributed fusion technology.For traditional distributed fusion,it relies heavily on track correlation,only achieves state fusion and loses target correlation and other high-level information,this paper studies the GLMB distribution for modeling the posterior distribution of complex multi-targets,and proposes GLMB distribution fusion to solve the fusion problem of the target posterior distribution;Aiming at the problem that the public information between sensors for optimal distributed fusion is difficult to calculate,the robust and sub-optimal generalized covariance intersection(GCI)fusion criterion is studied to avoid repeated calculation of public information between sensors.Finally,the feedback fusion structure is utilized to improve the local sensors tracking performance.