Research On Multi-target Tracking Algorithm Based On Multi-bernoulli Filter

With the continuous improvement of marine strategic position,tracking of multiple targets on the surface and underwater has become a new hot issue.Due to the complex marine environment and the development of target stealth and camouflage technologies,the detection of surface ship formations and underwater UUV formations by sonar equipment is facing huge challenges.At the same time,due to the uncertainty of the information obtained by the sensor,the multi-target tracking problem is more complicated.Based on the above background,research the passive multitarget tracking algorithm in the marine environment.First,the azimuth-only positioning method is briefly introduced;under non-linear observation conditions.The simulation compares the tracking performance of commonly used single-target filtering algorithms.The random finite set theory and the evaluation index of multi-target tracking performance are introduced.Secondly,the Multi-target Multi-Bernoulli(Me MBer)filtering algorithm based on random finite set is studied without considering the missed detection of the target.For the problem that the Me MBer filtering algorithm overestimates the number of targets when the missed target is considered,the posterior probability density of multiple targets is re-approximated,The Cardinality Balanced Me MBer(CBMe MBer)filtering algorithm is studied.For the problem that a single model is difficult to track accurately when the target is maneuvering,the IMM-CBMe MBer filtering algorithm combined with the interactive multi-model algorithm is studied.Finally,in view of the problem that the CBMe MBer filtering algorithm can only estimate the target state and cannot identify the target trajectory,the tag information is added to the target state,and the concept of a random finite set of tags is given.For the CBMe MBer filtering requires high detection probability and low clutter rate conditions,the multi-target posterior probability density is approximated as a generalized multi-Bernoulli distribution,the multi-target state is approximated by hypotheses with different weights,and the Labeled Multi-Bernoulli(LMB)filtering algorithm is studied.For the problem of unknown and time-varying target detection probability and clutter rate in practical engineering,the LMB filter algorithm with unknown detection probability and the LMB filter algorithm with unknown clutter rate are studied,and the effectiveness of the algorithm is verified by simulation analysis.