Research On Algorithms Of Tracking Non-Ellipsoidal Extended Targets Using Random Matrices

In recent years,due to rapid advances in sensor technology,extended target tracking is becoming more and more important in many military applications such as antimissile,warning system,guidance and battlefield surveillance.As the number of measurements obtained by the sensor increases,tracking algorithms have to estimate physical extension of an extended target in addition to its kinematic state.Non-ellipsoidal extended targets have received increasing research attention over the past decade.An extended target with irregular shape cannot be approximated by a simple geometric shape such as a circle,an ellipse or a rectangle.Based on Bayesian filtering,this thesis studies several promising shape estimation methods such as random matrix model,random hypersurface model and silhouette model.The emphasis is conducting research on tracking of non-ellipsoidal extended targets.The main contributions of this thesis are as follows:1.For the problem of non-ellipsoidal extended target tracking in maneuvering scenario,an improved algorithm based on adaptive random matrix model is proposed.Multiple ellipses are used to describe the shape of a single non-ellipsoidal extended target.The extension state of each sub-target is modeled by inverse Wishart distribution.In order to estimate the extended target shape accurately,it is necessary to establish an appropriate extension evolution model and a suitable measurement noise model.With prior estimated extension state and current measurement information,parameter adaptive approaches for both extension model and noise model are proposed.The simulation results show that the proposed algorithm improves the performance of extension state estimation while guaranteeing the accuracy of kinematic state estimation.2.For the problem of non-ellipsoidal extended target tracking in non-Gaussian systems,an improved algorithm based on elliptic random hypersurface model is proposed.In view of non-ellipsoidal extended target which can be assumed to be rigid body,a unified dynamic model is established to describe the kinematic states of all sub-targets.The proposed algorithm replaces the traditional linear measurement model with the pseudo measurement model.The shape parameters of the ellipse are estimated to avoid the treatment of the random matrix.The simulation results show the proposed algorithm is insensitive to the statistics of spatially distributed measurements.It is possible to estimate the target shape accurately even if few measurements are available per time step.3.For the problem of non-ellipsoidal extended target tracking in the presence of occlusions,an improved algorithm based on silhouette model is proposed.The measurement information obtained by sensors can be categorized into two types: the positive measurements and the negative measurements.The proposed algorithm incorporates the positive measurements assumed to obtain from the extended target and the negative measurements assumed to stem from elsewhere.The update step of the Bayesian estimator for extension state is modified to exploit information from both types of measurements.A gamma Gaussian inverse Wishart mixture implementation,which is capable of estimating the target extensions and measurement rates as well as the kinematic states of targets,is proposed.The simulation results show that target tracking performance is improved in situations with unexpected occlusions.The proposed algorithm is more robust and universal,compared to the previous work on tracking of non-ellipsoidal extended targets.