Research On Target Sensing Method In Complex Electromagnetic Space With Vector Bundle Model

Electromagnetic space is the main space of the electromagnetic activities,and the electromagnetic activities are the important approaches of the spatial information acqui-sition.In the initial stage of radio communication and sensing,it is easy to extract the embedded information in signal,as the pure electromagnetic environment.However,nowadays,the gradually complex electromagnetic environment has changed the structure and distribution of electromagnetic space,the classical characteristics of electromagnetic signal are overwhelmed in the background.As the gradual complexity of the electromag-netic space,the previous cognition of electromagnetic space and the target sensing method are invalid anymore.So,there is a challenge in the spatial information acquisition method and processing.In this question,the basic problems are the structure of the space,the law of state change and the relation between the space state and object.For resolving the target sensing problem in complex electromagnetic space,the thesis researches on the complex electromagnetic space model,target detection and target localization,and proposes the satisfied novel theory and methods for providing the theoretical foundation and practical guidance of spatial information acquisition.The first chapter introduces the concept of electromagnetic space and the problems in space modeling and target sensing.Moreover,the way of target sensing in complex electromagnetic space is compared with the radar traditional approach.In addition,the references are summarized.In the second chapter,the vector bundle model of complex electromagnetic space is proposed.And,the related abstract concepts in the information geometry and differential geometry are introduced in the understandable way.Furthermore,based on the vector bundle model of complex electromagnetic space,the spatial change is modeled as the section distance function,that provides the foundation for the following chapters.The third chapter researches on the target detection problem in the complex electro-magnetic space.Under the background of detection problem in electromagnetic space and the vector bundle model,the core problem of target detection is the information geometry based unknown signal detection problem.In this chapter,the performance of geometry detector for unknown signal is discussed.In the discussion,the statistical distribution of test statistic is deduced,the relation between detection performance and signal character-istics is analyzed,and the recursive method for threshold is provided.Then,based on the section distance function,the method,which utilizes the perturbation to detect target,is proposed.In the proposed method,the l_k-norm is used to obtain test statistic for target detection in complex electromagnetic space.Finally,the proposed method is improved in the scene with cooperative transmitter and analyzed in multi-target detection,that expands the range of application of the proposed method.The forth chapter researches on the target localization problem in the complex elec-tromagnetic space.Based on the vector bundle model of complex electromagnetic space and the section distance function model of space change,the localization problem is trans-formed to the mapping acquisition problem.Based on such point,this chapter presents the data-driven target localization framework for the complex electromagnetic space.In particular,because of the simple structure,the radial basis function network is selected;the deep neural network is also selected as its advantages in regression.The performance of this two sorts of networks are studied in the proposed localization framework.Because of the shortcoming of radial basis function network and deep neural network,the two stage shunt-wound radial basis function network system and mutually inverse networks are proposed to get better performance.The two stage shunt-wound radial basis function network system is able to fuse two radial basis function network to get better results,by the two stage network system.The mutually inverse network utilizes a radial basis func-tion network to assist training,that increases the data utilization ratio and improves the localization performance.The fifth chapter provides the summary of the thesis and points out the further study topics.The contribution of this thesis provides the theoretical foundation and practical guid-ance to the target sensing issues in electromagnetic space,and offers the valuable reference to the further studies and applications.