Research On Multi-target Tracking Method Combined GPF And Data Association Algorithms Based On Passive Sonar Systems

Target location and tracking is the first step of naval operations.The improvement of our maritime superiority need positioning and tracking skills more accurately at the beginning.The positioning mode of active sonar is a kind of active positioning which achieve its goals by send signals to external environment.A fatal flaw of active positioning mode is this method may be expose ourselves.Passive sonar achieve its positioning function by receiving signal which transmitted from the enemy.This method called bearings-only target motion analysis which use the bearing information of target radiation source signal to obtain target location and trajectory.Because at the same time of positioning and tracking,our submarine remain silence to protect ourselves,so that this analysis method can be widely used in marine warfare.As the targets maneuvers more and more complex,the accurate tracking of multiple maneuvering targets has a better practical value.For maneuvering targets,the nonlinear filter can be more suitable to for the situation when the target maneuvers.Particle filter is not too restrictive to the target system and the noise properties than other nonlinear kalman filters which have some move model and noise properties limitations.Therefore,particle filter can handle arbitrary nonlinear models and arbitrary noise distributions.The applicability of the particle filter algorithm to the nonlinear characteristics of underwater acoustic noise is analyzed.Gaussian particle filter becomes more simple and more easily to realize because it is not need re-sampling procedure,and this step precisely solved the problem of particle scarcity and particle degradation which exists in particle filter.Multi-target tracking is based on single target tracking,and it need to consider a vital question than single target tracking whether there has mutual influence between each track.The target trajectory is prone to track merge phenomenon when target is overlapping movement or target moves in parallel.The joint probability data association expresses the relationship between each observation trace and the track with the probability of association,which can be used to track a known number of goals without any prior information about the target and the clutter.This paper analyses a view of the measured point associated with all trajectories and improves the classical joint probability data association method,simulation verifies that some effect is achieved.The last chapter of this article analyzes the basic processing flow of the multi-target tracking system from the view of trajectory management.Data preprocessing,track start,track association,point supplement and track demise are considered in this system.Finally,an effective tracking is achieved by processing a measured bearing time record combined with nonlinear filtering and track overlapping analysis.Simulation results verifies that the system can track multi-target effectively based on bearings-only target motion analysis.