| The noncooperative bistatic radar belongs to a passive detection system that uses noncooperative radar as the illuminator and carries out target detection based on the working method of bistatic radar.It is covert,has low costs of construction and operation,and it has the potential to counter stealth technology.Aiming at the problem of target tracking of noncooperative bistatic radar,this thesis carries out the research work of bistatic radar nonlinear filtering algorithm,multi-target tracking under low detection probability and high clutter rate,and target tracking with unknown radar illumination information.The second chapter introduces the system of noncooperative bistatic radar,and establish target tracking models.The system structure of noncooperative bistatic radar,signal processing flow,and target tracking models are introduced.By comparing with traditional active monostatic radar,bistatic radar and passive radar using commercial illuminator,the challenging issues faced by noncooperative bistatic radar target tracking are introduced: measurement nonlinearity,low detection probability and high clutter rate,and unknown radar illumination information.The third chapter studies the nonlinear filtering algorithm of bistatic radar.Firstly,the development of bistatic converted measurement Kalman filter and three classical nonlinear filtering algorithms including extended Kalman filter,unscented Kalman filter,and particle filter are introduced.Then the problems of traditional bistatic converted measurement Kalman filter are analyzed,and bistatic decorrelated unbiased converted measurement and the corresponding Kalman filter are proposed using Taylor series expansion.Finally,computer simulations illustrate unbiasedness and covariance consistence of bistatic decorrelated unbiased converted measurement and the superiority on accuracy of target state estimation and covariance consistence of the corresponding Kalman filter.The fourth chapter studies multi-target tracking under low detection probability and high clutter rate,using mechanical scanning radar as the illuminator.Firstly,the probability hypothesis density filter and its sequential Monte Carlo implementation are introduced.The problems of probability hypothesis density filter under low detection probability and high clutter rate are analyzed: target loss,large error of target state estimation,and too many false targets.Then the online target distinguishment based probability hypothesis density filter is proposed,and the main features are a new method of track identification and state estimation,the survival probability dependent on the target state,and online distinguishment between real target and false target using sequential probability ratio test.Computer simulations illustrate that the online target distinguishment based probability hypothesis density filter can avoid target loss caused by continuous miss detection,eliminate false targets caused by high clutter rate,and deal with the problem of low detection probability and high clutter rate in noncooperative bistatic radar target tracking effectively.Finally,the field experiment verifies the effectiveness and practicability of the proposed algorithm.The fifth chapter also studies multi-target tracking under low detection probability and high clutter rate,using mechanical scanning radar as the illuminator.To apply multi-Bernoulli mixture filter to nonlinear scenes,its sequential Monte Carlo implementation is proposed,including particle system description,Gibbs sampling,ellipsoidal gating,target/hypothesis pruning,track identification and state estimate.Then the problems of multi-Bernoulli mixture filter under low detection probability and high clutter rate are analyzed: track breakage,track termination lag,and too many false targets.The offline target confirmation based multi-Bernoulli mixture filter is proposed,and the main features are the survival probability dependent on the target state and real targets offline confirmation using hypothesis test.Computer simulations illustrate that the offline target confirmation based multi-Bernoulli mixture filter can deal with the problem of low detection probability and high clutter rate in noncooperative bistatic radar target tracking effectively,and outperforms than the online target distinguishment based probability hypothesis density filter under large process/measurement noise.Finally,the field experiment verifies the effectiveness and practicability of the proposed algorithm.The sixth chapter studies target tracking with unknown radar illumination information,using phased array radar as the illuminator.Firstly,target tracking model with unknown radar illumination information are established.Then for single target tracking with known radar illumination information,the intermittent Bayesian filter is proposed,in which different correction equations are used to update the target state according to whether the target is illuminated.For single target tracking with unknown radar illumination information,the soft Bayesian filter and the hard Bayesian filter are proposed,in which the second state variable is constructed to represent whether the radar illumination is present,both the target state and the unknown radar illumination information are estimated in Bayesian filtering recursion framework.The main difference between the soft Bayesian filter and the hard Bayesian filter is whether to make a decision on the radar illumination information explicitly.Finally,computer simulations demonstrate that the proposed algorithms can realize the joint estimation of target state and illumination information of phased array radar,and can effectively track single target with unknown radar illumination information.The seventh chapter summarizes this thesis and looks forward to the next research work. |
