Research On The State Estimation Method For Perceived Targets Of Passive Sensors

In recent years,the development of modern anti-radiation missile technology and target electromagnetic stealth technology has brought a huge threat to active radar.Due to the better concealment of passive detection technology,multi-target tracking technology based on passive sensor systems is increasingly receiving attention.Using angular measurement information such as azimuth and elevation angle to achieve pure-angle tracking is the main direction of multi-passive sensor tracking.For the passive sensor system,this paper adopts an improved multi-dimensional distributed data association method and a multi-target tracking method under the framework of random set theory.Finally,in the three-dimensional space with clutter environment,multi-objective state estimation is realized.The main research content of this article is as follows:First,based on Bayesian theory,Kalman Filter(KF),Extended Kalman Filter(EKF),Unscented Kalman Filter(UKF)and Particle Filter(PF)are studied.It lays the foundation for the non-linear system target tracking by combining the nonlinear filtering method with the random set theory.Secondly,the traditional multi-target tracking method needs data association,which will cause the problem of combinatorial explosion of computation.The multi-target tracking method under the Random finite set theory framework is studied,and the target motion model and sensor observation model are established.For Bayesian filtering in the Random finite set framework,there are some problems that can not be solved due to problems such as set integrals.Probability Hypothesis Density(PHD)filters is introduced.,and the essential relationship between PHD filters and Bayesian filters is discussed.Thirdly,this paper studies passive sensor measurement data association and fusion methods.To solve the problem of excessive calculation of the traditional multi-dimensional allocation algorithm,an improved 3-D allocation algorithm is proposed,which can guarantee the accuracy and reduce the amount of calculation.The array multidimensional data fusion method and cross-location fusion method for passive multi-sensor systems have also been studied.This paper uses these two methods to fuse the correlated measurement data to obtain the input amount of the tracking filter.Subsequently,for the problem that the PHD filter cannot obtain a closed solution,two implementations are introduced: Gaussian mixture PHD and particle PHD.In order to realize target tracking using Gaussian hybrid PHD filtering in nonlinear systems.,the EKF and UKF filtering algorithms were combined with the GM-PHD filter.The simulation experiments based on the passive sensor system were designed to prove the effectiveness of the two filters.The results were analyzed.Finally,a multi-model approach is introduced for the more mobile targets and combined with a linear GM-PHD filter to propose an IMM-GM-PHD filter.A simulation experiment was designed to verify the effectiveness of the IMM-GM-PHD filter in a linear Gaussian system.This paper analyzes the results.The multi-model method was also combined with the particle PHD to propose the IMM-SMC-PHD filter.The simulation experiment based on the passive sensor system is designed,and ultimately realize the multi-objective state estimation of multi-passive sensors in 3D space clutter environment.