Research On Passive Radar Target Detection,Tracking And Classification

Passive radar has the abilities of anti-interference,lowcost and easy to deploy.It has attracted a lot of attention due to the prospects of application in reconnaissance,supervisory and stealth/low-altitude target detection.In this dissertation,an in-depth study of the topic of detection,tracking and classification of targets is carried out.Several algorithms and models are proposed,and their performance is examined through theoretical analysis or simulation experiments.The main contributions are summarized as follows:1.The problem of detection and tracking of multiple targets with existence of registration biases is addressed.The dynamics of registration biases are firstly modeled as the first-order Markov process,and then augmented into the target state vector.Based on the multi-object multi-Bernoulli filter,an algorithm of joint target detection,tracking and sensor registration is proposed.The particle filtering implementation of the proposed algorithm is also introduced.2.The problem of detection and tracking of a very low observable target based on passive radar is addressed.Based on the principle of probabilistic data association,the batch log-likelihood function of passive radar is obtained in order to integrate target measurements throughout sampling time.The estimated target state is found by maximizing the batch log-likelihood function.In the algorithm level,the particle swarm optimization method is introduced to find the global maximum of log-likelihood function,then the observation guided particle initialization method is proposed to promote the computational efficiency.In the implementing level,the GPU is employed to carry out the proposed method in parallel,and the implementation strategy is designed by carefully considering the properties of the log-likelihood function.As a result,the problem of fast detection and tracking of a very low observable target is solved.3.Based on the finite set statistics framework,a sub-optimal likelihood function of detection and tracking a target of single frequency network passive radar is proposed.The optimal likelihood function is re-constructed as the sum of sub-likelihood function related to every transmitter.The target measurements produced by other transmitters are considered as clutters when considering specific sub-likelihood function.The advantage of the decomposition is that the association ambiguities between measurements and transmitters can be avoided,and the proposed likelihood function achieves substantial improvement on computational efficiency with just little accuracy loss.4.A GCI fusion based algorithm is proposed to address the problem of multitarget detection and tracking using single frequency network passive radar.In the proposed method,every transmitter is regarded as a virtual node.At first,the local posterior distribution of each virtual node is computed,where the targets measurements produced by other virtual nodes are regarded as clutters.Then the GCI method is employed to fuse the local distributions to the global distribution.Compared to the existing methods,the proposed method has lower computational load and better scalability.5.An algorithm is proposed to address the problem of joint detection,tracking and classification of multiple targets with passive radar.In the proposed algorithm,target class information is included into the multi-model multi-object multi-Bernoulli filter.By utilizing the prior information between target class and kinematic model set,the performance of target detection and tracking is promoted.By fusing the class information provided by target model set and class measurement provided by passive radar,the performance of target classification is promoted in turn.