Research On Key Techniques Of Time-frequency Difference Passive Localization Based On Signal-level Fusion

Passive localization is an important part of the field of the battlefield electronic reconnaissance,capable of obtaining the coordinate position of the radiation source target by passive means.It plays an important role in seizing the right to battlefield information and occupying the dominant right to war.In recent years,with the continuous expansion of the type of radiation source and the increasing capability of the front-end reconnaissance sensor,the radar,communication,data chain and other radiation sources as well as external radiation source signals from the same target can be effectively detected and used for passive localization.The traditional solution is to fuse the target localization results of different signals at the data level,which does not improve the localization accuracy significantly.As a result,based on the idea of fusing multiple radiation source signals,three key issues,namely,estimation of localization parameters for signal-level fusion,network localization of fused signals,and optimization of station placement under constraints,were studied in depth in this paper.It has provided support for improving the performance of localization algorithms and refining the methods of localization technology.The main work of the dissertation is as follows:1.In the research on localization parameter estimation by signal-level fusion,a secondary coherent accumulation time-frequency difference estimation method based on signal fusion was proposed for multi-signals in the form of pulses;a two-segment sampling accumulation timefrequency difference estimation method based on signal fusion was proposed for multi-signals in the form of pulses and continuous waves.In the secondary coherent accumulation method based on signal fusion,the multi-pulse information in the fused signals was united,and two single signals were used as the reference to accumulate the fused signals twice,and the peak position of the cross-ambiguity function in the secondary accumulated signals was used to estimate the time-frequency difference.Simulation experiments showed that the proposed method improves the accuracy of TDOA parameter estimation by 57.7% compared to single signals and improves the localization accuracy by 48.1%compared to data-level fusion methods in the setup scenario.In the two-segment sampling accumulation method based on signal fusion,the signal was accumulated at the sampling rate of the pulse signal,and the remaining portion corresponding to the continuous signal was shifted and then accumulated at the original sampling rate,and the time-frequency difference was estimated by the peak position of the cross-ambiguity function.Simulation experiments showed that the proposed method can improve the estimation accuracy of both TDOA and FDOA compared to the parameter estimation method of single signal.2.In the research on target localization with fused signals,to address the problems of accuracy,robustness and complexity in traditional methods,a localization method based on recurrent neural networks was proposed,and the method was improved by using the error model of the time-difference data.In the localization method based on recurrent neural network,the localization equation was established for the target activity region,the data pairs of target position and measurement value were obtained,the recurrent neural network was trained,which was used to locate real-time targets appearing in the region.Simulation experiments verified that the localization method based on the recurrent network model has a higher localization accuracy than the traditional method,and that no unsteady point exists in the target area,and the efficiency of localization operation is significantly improved over the traditional methods.On this basis,for the problem of noise in the target region,the error model of time-difference data was established,and the network localization method was improved by constructing data error terms to train the network.Simulation experiments showed that the localization method of the neural network improved by using the error model has a higher localization accuracy in the noise environment.3.In the research on localization station placement with constraints,a geometric station placement method based on the classification of angular critical value was proposed for the issue of signal beam angle restriction;a grey wolf optimization station placement method based on the target region error model was proposed for the issue of target region range restriction.The geometric station placement method based on the classification of angular critical value firstly gave the influence of the angle variable of the station on the localization error in the threestation localization scenario.Then,the theoretical error function was used to derive the angular critical value,the four-equivalent angle of the two stationary stations and the five-equivalent angle of the beam.At last,the relationship between the positions of the three stations,the boundary values of the beam angle range,and the derived angular critical values was utilized to give the geometrical station placement methods in different cases.Simulation experiments verified that the proposed geometric method based on the classification of angular critical value can minimize the localization error.The grey wolf optimization station placement method based on target region error used the presence probability of the target signal in the region and the estimation accuracy of the target localization parameters to construct the region error model based on the position probability;the overall error matrix under the model was taken as the objective function,and the grey wolf algorithm with improved parameters was used for the optimization solution.Simulation experiments verified that the grey wolf optimization method under the model can quickly converge to the optimal station position,and has a smaller regional error for target localization.