Research On Algorithm Of Single Target Tracking Based On Extended Kalman Filter

With the rapid development of science and technology,especially the rapid advancement of aerospace technology,several countries have developed high-speed and high-mobility aircraft.These targets have the characteristics of high sound speed,large-scale maneuvering,nonlinear system model and time-varying parameters.In the target tracking system,the accurate estimation of the target state is especially critical,and the actual system model is mostly nonlinear,which puts higher requirements on the nonlinear filtering algorithm.In this paper,the extended Kalman filter is studied.The main work of the thesis include:Several classical nonlinear filtering algorithms in target tracking are analyzed,including Extended Kalman Filter(EKF),Unscented Kalman Filter(UKF),Particle Filter(PF),Current Statistical Model(CSM)and Interacting Multiple Model(IMM)algorithms.The UKF uses the unscented transformation to deal with the mean and covariance of nonlinear systems,but there are problems that rely on empirical selection parameters and that the filter is prone to divergence when the state dimension is high.PF can be used in non-Gaussian systems,however,to improve the filtering accuracy,it is necessary to increase the number of particles,and the real-time is poor,and there is a problem of particle degradation.The CSM takes into account the effect of acceleration on state estimation,but when the acceleration limit is set too large,the prediction covariance is large,which reduces the tracking effect of the filter on weakly maneuverable or non-maneuvering targets.IMM is a multi-model switching algorithm that overcomes some shortcomings of the single-model algorithm,but the accurate selection of the dependent model set and the Markov chain transition state matrix is not easy to implement.Considering that the tracking system should meet the requirements of easy engineering implementation and real-time tracking,it is important to determine the key research EKF algorithm,which has the characteristics of simple structure,easy implementation and good real-time.It is a hotspot and effective method for studying nonlinear filtering algorithms now and in the future.Aiming at the problem that the prediction covariance is too large deviation from the true value of the target maneuver,which leads to inaccurate EKF tracking,an improved extended Kalman filter algorithm is proposed.The improved algorithm adopts the EKF algorithm framework,and uses the idea of adaptive fading filtering.According to the filter divergence criterion,a new adaptive fading factor is proposed and introduced into the calculation of prediction covariance.Using the new fading factor can overcome the problem that the traditional fading factor does not take into account whether the filter is currently diverging or converging when it is ascertained.Improving the calculation accuracy of the fading factor.Using the Levenberg-Marquardt method to predict the association.The variance is iteratively optimized to gradually fit with the real covariance,and then the filter gain is adjustedin real time to improve the accuracy of the filter tracking and ensure the stable convergence of the algorithm.The simulation experiment results of tracking maneuvering targets in three-dimensional space show that the tracking accuracy of the improved algorithm is higher than that of the traditional EKF,UKF and PF algorithms,and the tracking accuracy of CSM and IMM is not much different,and the real-time is better than UKF,PF,CSM and IMM algorithms.The effectiveness of the improved algorithm is proved.