Multi-target Tracking Based On Distributed Data Fusion

With the continuous development of the ocean and the gradual improvement of its strategic position,the ability of multi-target tracking has increasingly become a major factor restricting various underwater activities.However,due to the high clutter of underwater environment itself,slow underwater acoustic propagation speed,low signal frequency band and other factors,the measurements not only contain a small amount of real target information,but also contain a large number of clutters.In addition,how to use the measurements of multi-platform to obtain more reliable target tracking trajectories is also a common problem.Under this background,this paper mainly studies the multi-target tracking methods in clutter environment based on distributed data fusion theory through three parts: contact data preprocessing,multi-target tracking at platform-level and track fusion at system-level.The methods are simulated and analyzed by MATLAB and verified by trial data.Firstly,the preprocessing technology of contact-level information is studied.A multiplatform detection data model is constructed according to the properties of contact-level data which contains the measurements of the real target and clutter.Then,the space-time errors of the measurements are registered by pre-processing steps such as conversion between different coordinate systems,space deviation calibration based on real-time quality control method or least square method,and time registration based on virtual fusion method or interpolation extrapolation method.Secondly,multi-target tracking method at platform-level is studied.Starting with the track initiation algorithms such as logic method and Hough method,the initial state of the tracking target is obtained.Then,kalman filter and probabilistic association algorithm are used to track single target without or with clutter.Then joint probability data association algorithm is used to track multi-target in high clutter density.Then the joint probabilistic data association algorithm is used to realize multi-target tracking under high clutter density.Finally,a multi-target tracking model based on measurements is proposed to achieve multi-target tracking in time-varying number scenarios.Thirdly,the fusion processing of multi-platform tracking trajectory is studied.Based on the distributed data fusion model,the track of the local platform is correlated by the weighted correlation algorithm or Hungarian algorithm,and then the track output at the system-level is obtained by the fusion algorithm of covariance convex combination and covariance intersection respectively according to the correlation of the local estimated covariance.Finally,the sea trial data are processed.The multi-target tracking algorithm based on measurements is used to obtain the platform-level track,and then the system-level track is formed by covariance intersection algorithm.The multi-target tracking of distributed multiplatform without target prior information is validated.