| Moving target tracking techniques focus on the data processing for real-time target dynamic state estimation and track formation based on sensor measurements,which has been widely used in military and civil fields,such as air-to-air and ground-to-air flight surveillance systems.In practical target tracking systems with different task requirements,platform amounts,and platform features(stationary or moving,ground-based or space-based),several issues still need to be addressed,including the complex nonlinear filtering in air-to-air target tracking systems caused by a series of coordinate conversions,the incompletely observable state in single 2-D radar 3-D target tracking systems,and the measurement time-offset estimation in multisensor target tracking systems.Focusing on these issues,effective solutions are designed by implementing modern estimation theory and information fusion techniques,and the corresponding simulation experiments are carried out to verify their efficiency.The main contributions of this thesis are as follows:1.An Unbiased Converted Measurement Kalman Filter in Earth-Centered Earth-Fixed(ECEF)coordinates(E-UCMKF)is proposed for dealing with the measurement conversion bias in the existing converted measurement filter for air-to-air target tracking,which is caused by ignoring platform navigation errors.In the proposed algorithm,airborne radar measurements are converted from spherical coordinates to ECEF coordinates without bias,taking account of noisy platform navigation measurements,and then the Kalman filter is implemented for target state estimation.When range-rate measurements are available to the processor,a Sequential Unbiased Converted Measurement Nonlinear Filter with Range-rate Measurements(E-SUCMNFw R)is developed,which constructs a pseudo measurement to decorrelate the range-rate measurement noise from the unbiased converted measurement noise,and updates the state sequentially by the E-UCMKF and a modified Cubature Kalman Filter(CKF).Simulation results validate that the consistency of the proposed unbiased measurement conversion outperforms traditional measurement conversion approaches and that the proposed E-UCMKF and E-SUCMNFw R outperform the corresponding conventional algorithms.2.A Height-Parameterized Extended Kalman Filter with Probabilistic Data Association(HP-PDA-EKF)algorithm is proposed to resolve the inconsistency between the initial filter state and the corresponding standard deviation in existing height-parameterized 2-D radar 3-D target tracking algorithms and to deal with the target tracking problem with false alarms and missed detections.Both the PDA algorithm and the nonlinear filter are implemented in the proposed algorithm so that it is applicable for target state estimation in the presence of false alarms and missed detections.Moreover,the standard deviation of initial target state is approximated based on the Monte Carlo method to ensure the consistency between the initial state and the corresponding standard deviation.The Posterior Cramér-Rao Lower Bound(PCRLB)for the 2-D radar 3-D target state estimation is derived to provide the theoretical lower bound on the estimation accuracy.Simulations in different scenes with different initial slant range between the target and radar are carried out.The results of tracking performance comparison show that the proposed algorithm outperforms the existing algorithms.3.To improve the observability of ground-to-air 2-D radar 3-D target tracking systems,the waypoint information is used for target tracking and a Height-Parametrized Cascaded PDA Filter(HP-CPDAF)is proposed.In this algorithm,a set of sub-waypoints along the airway across the prior waypoints are interpolated as extra measurements,which are then associated with the initialized track to update the state after a cycle of HP-PDA-EKF.Taking account of the waypoint information,the PCRLB expression is derived.Sets of simulations are carried out to analyze the state estimation accuracy with different waypoint accuracies and sub-waypoint amounts,providing a reference for the application scope and parameter setting of the proposed algorithm.Simulation results also demonstrate that the introduction of waypoint information can effectively improve the target tracking accuracy.4.Motivated by pseudo-measurement methods for spatial error registration,a pseudo-measurement based multisensor measurement time-offset estimation algorithm is proposed to deal with the measurement time offset in multisensor non-maneuvering target tracking systems.Based on either measurements or local tracks obtained from two sensors,the pseudo-measurement equation of time offset is derived for Constant Velocity(CV)target tracking,and the observability of both absolute time offsets and relative time offsets between sensors is analyzed.Since the absolute time offset is unobservable,an Unbiased Converted Measurement Kalman Filter-Recursive Least Squares/Kalman Filter(UCMKF-RLS/KF)algorithm is designed for the constant/time-varying relative time-offset estimation.The PCRLB expression is derived to quantify the relative time-offset estimation performance.Simulation results for single CV target tracking show that the Root Mean Square Error(RMSE)values of the relative time-offset estimates approach the theoretical lower bound.5.To deal with the degradation of estimation accuracy and correct association rate caused by measurement time offsets,a distributed multisensor multitarget tracking algorithm with time-offset registration is proposed for distributed multisensor multiple CV targets tracking in clutters.In the global processor,the equivalent measurements are constructed using the time-biased local tracks based on the inverse Kalman filter and the pseudo-measurements are calculated using the equivalent measurements.Then,the recursive least squares estimator and the Kalman filter are implemented to jointly estimate the state in space and time domains,respectively.Finally,a framework of distributed multisensor multitarget tracking with time-offset registration is presented,where the time-varying relative time-offset estimation and compensation,equivalent measurement to global track“association,and global track update are included.Simulation results show that the proposed algorithm can effectively estimate and compensate the relative time offset and finally provides more accurate global tracks.6.Focusing on the measurement time offset in multisensor maneuvering target tracking systems,a measurement time-offset estimation algorithm in multisensor maneuvering target tracking systems is proposed.Firstly,the expressions of time-offset pseudo-measurement equations for the Constant Acceleration(CA)and Coordinated Turning(CT)target tracking are derived,respectively,and the corresponding observability analyses are performed.Since the absolute time offset is unobservable,an Unbiased Converted Measurement Kalman Filter-Iterative Least Squares/Kalman Filter(UCMKF-ILS/KF)algorithm is designed for the constant/time-varying relative time-offset estimation in non-maneuvering target tracking systems.Furthermore,the Interacting Multiple Model Two-stage Filter(IMM-TF)is designed for estimating the relative time-offset in maneuvering target tracking systems.Then,the proposed algorithm using two sensors is extended to multisensor cases to obtain the minimum-bias time-offset estimates.Finally,the relative time-offset estimation PCRLBs for all cases are derived.Simulation results show that the RMSE values of time-offset estimates in the case of CA,CT,or maneuvering target tracking are close to the corresponding theoretical lower bounds and within their 95% confidence intervals. |
