Research On The Underwater Multiple Target Tracking Based On Information Fusion

With the rapid development of information technology and computer technology,exploiting and making full use of the ocean resources has gained more and more attention by many scholars and organization all over the world in these years.Because of the extensive application prospects on the military and civilian fields,underwater target tracking problem has gained wide concern by the society.How to discover offshore and underwater targets timely,and to track targets accurately become one of the important hot topic in the filed of domestic and foreign.There are still many annoying problems need to be resolved at present,such as the convergence,robustness,accuracy and real-time performance of the filter algorithm,the measurement uncertainty and fuzzy data problems of multiple targets tracking,the nonlinear characteristic of tracking system,and the uncertainty problems of model parameters.The aim of this thesis is to improve the target tracking performance of underwater sonar system.Taking advantage of the two key technologies,recursive Bayesian filter technology and data association method,underwater target tracking problems based on the information fusion theory are investigated deeply.Combining the actual demand of underwater target tracking system,underwater single target tracking problem based on multiple passive information,active multiple targets tracking problem based on probabilistic multiple hypothesis tracking(PMHT)algorithm,bearing-only multiple targets tracking problem under dense clutter and multiple targets tracking based on range-Doppler measurements in dense clutter are studied in this thesis.The research results can provide technical reference for complex target tracking scenario of sonar system,and it has important theoretical and practical significance for improving the tracking capability of moving target in sonar tracking system.The main contents and innovations of this thesis are summarized as follows.1.Combining the target's state model and measurement model of underwater passive target tracking system based on multiple information process,this thesis firstly studies the target's state-observability problem for single passive observer target tracking based on bearing-only measurement information.What's more,by introducing Doppler frequency measurement information,the passive tracking system can be observable on the condition of without the observer's maneuvering moving.First,the thesis shows that only if the passive observer maneuvering moving,the targets' state is observable.On the other hand,for the multiple observer environment,the targets' state is observable if only the target does not move on the line of the multiple observers.Second,the observability problem for single passive observer target tracking based on bearing and Doppler information is studied.By linearizing the nonlinear measurement equations,the pseudo measurement can reduce the computational burden.It shows that if only the target does not move toward the single observer radially,the targets' state is observable.Last,we investigate the performance of the recursive extended Kalman filter(EKF)algorithm on the single target tracking based on bearing-only measurements and bearing and Doppler measurements for the cases of single stationary observer,the single maneuvering observer and multiple observers,respectively.The simulation results show the suitability and effectiveness of the EKF algorithm.2.To solve the data association problems of underwater multiple targets tracking on dense clutter environment and incomplete detection environment,the thesis derived the homothetic probabilistic multiple hypothesis tracking(HPMHT)algorithm.To solve the data association problems between the measurements and targets under the underwater dense clutter environment,combining the idea of the expectation-maximization algorithm and the PMHT algorithm,the HPMHT algorithm decomposes the measurements information from the same target to a multiple measurements.What counts is that these measurements which come from the same target have the same mean and different covariance with the specific target.The simulation results show the position root mean square errors of multiple crossing moving targets and uniform linear neighboring moving targets for HPMHT algorithm are small,and the HPMHT algorithm can greatly increase multiple targets' correct tracking rate with short calculation time.3.The PMHT algorithm's performance is sensitive to the targets' initial state in most situations.Combining the deterministic annealing technology and the idea of HPMHT algorithm,the thesis derived the deterministic annealing homothetic probabilistic multiple hypothesis tracking(DA-HPMHT)algorithm.On the bases of the deterministic annealing procedure and HPMHT algorithm,and by introducing the annealing factor ? in the conditional probability mass function of the targets,the DA-HPMHT algorithm can expand the relevance's between the measurements and targets,and thus improves the calculation accuracy of target's posterior association probability.The Monte Carlo simulation results show that the DA-HPMHT algorithm has good performance under dense clutter environment for uniform linear crossing motion targets,maneuvering crossing targets and uniform linear neighboring motion targets,and the algorithm can meet the computer real-time request,which show the effectiveness of the proposed algorithm.4.The underwater bearing-only multiple targets tracking problem under dense clutter environment is discussed,and the multiple sensors PMHT algorithm suitable for bearing-only target tracking conditions is proposed.Different from the underwater bearing-only single target tracking problem,due to the data fuzzy problem between multiple targets and measurements in the case of bearing-only multiple targets tracking,the target tracking system need to handle not only the nonlinear filter problem but also data association problem.The data association process between the targets and measurements are independent of each other for the bearing-only PMHT algorithm,which deal with the data fuzzy problem between the targets and measurements by introducing a associated variable.Moreover,the sequential method is used to handle the data fusion problem for multiple sensors.The simulations prove the validity of the PMHT algorithm in the application of bearing-only multiple target tracking for multiple stationary passive sensors and maneuvering single passive sensor under dense clutter environment,and the simulations show that the tracking performance is sensitive to the physical position of the passive sensors.5.For the multiple targets tracking problem based on the measurements of range and Doppler information under dense clutter environment,the thesis proposed new extended Kaman filter(EKF)-based PMHT(PMHTe)algorithm and the unscented Kaman filter(UKF)-based PMHT(PMHTu)algorithm.The system locates targets according to ellipse in the two dimensional space and ellipsoid in the three dimensional space,which might generate a lot of “ghost” easily.The appearance of ghost makes it difficult to deal with the data fuzzy problem for multiple stations sonar tracking system.To deal with the data association problem between transmitters and measurements,the PMHTe algorithm and PMHTu algorithm introduce a new association variable for measurements and transmitters.And the simulation results show that the proposed algorithm has good tracking performance for multiple targets tracking based on the range and Doppler measurements under dense clutter environment,and the computational complexity is low.