Research On Adaptive Tracking And Information Fusion Of Emitters In Complex Electromagnetic Environment

The complex electromagnetic environment requires the spectrum sensing system to have ability to process signals and information from a large number of and densely distributed emitters.However,traditional reconnaissance processing methods may have problems,such as multi-signal detection errors and large measurement errors in such environment,which is difficult to meet the requirements.In view of this problem,this dissertation introduces the theory of random finite set,which can unify the mathematical description of the phenomena of the appearance,survival,disappearance,missed detection and false alarm of the emitter target,and better target tracking performance is obtained.This dissertation studies the problems involved in emitter reconnaissance,such as downfrequency data exchange,emitter classification,and robust emitter tracking,etc.The main work and contributions are described as follows:1.In order to reduce the communication load between receiving stations of the emitter reconnaissance system,the event-triggered strategy of information transmission between the receiving stations and the estimation problem of the multi-emitter state under the event-triggered strategy are studied.This dissertation proposes joint-transmission and independent-transmission event-triggered strategies,so that each receiving node can only transmit target information with sufficient information gain.Then,for the problem of distributed multi-emitter state estimation under the event-triggered strategy,the consensus labeled multi-Bernoulli(LMB)filter based on the distributed event-triggered strategy is proposed,and its implementation method with low computational complexity is given.2.The multi-receiver emitter target joint detection,tracking and classification(JDTC)is studied to improve the collaborative efficiency of multiple reconnaissance receivers.This dissertation extends the statistical description of the Bernoulli random finite set,and the emitter state is uniquely characterized by existence probability,class probability,classrelated model probability,and class&model-related state probability density.Then,based on the centralized and distributed fusion rules,the centralized and distributed processing fusion algorithms for the JDTC of the emitter target are derived.Both fusion algorithms improve the tracking and classification performance of emitter reconnaissance.3.In order to improve the detection,tracking and classification performance of multiemitter,a fusion method that can be used for the distributed multi-emitter JDTC is studied.This dissertation uniquely characterizes each labeled potential target by existence probability,class probability,class-related model probability,and class&model-related state probability density,the LMB distribution that describes the multi-emitter state is obtained,and the multi-emitter JDTC LMB filter is improved.Then,based on generalized covariance intersection(GCI)and minimum information loss(MIL)criteria,a fusion method for the augmented LMB distribution is further proposed.The global posterior distribution is obtained by fusing the posterior distribution of each local receiving node,and finally the distributed multi-emitter JDTC method is formed.Compared with the existing algorithms,the proposed algorithms in this dissertation have the advantages of small computation and strong scalability.4.In order to improve the robustness of the multi-emitter tracking algorithm under the condition of unknown noise distribution parameters,joint detection,tracking and parameter estimation of multi-emitter is studied.This dissertation augments the covariance matrix of the measurement noise distribution into the target state vector,so that the joint distribution of the augmented state vector conforms the Gaussian Inverse Wishart Mixture(GIWM)distribution.Then,the variational Bayesian(VB)technique is used to derive the same form of closed solution as the predicted distribution.The VB lower bound is used to construct a pseudo-likelihood model of the augmented state.Based on standard LMB,a fast LMB adaptive filtering algorithm based on superpositional measurement model is proposed,which improves the robustness and fault tolerance of the algorithm for unknown noise distribution parameters.