Distributed Rader Signal Sorting And Recognition

In radar electronic reconnaissance,signal sorting and identification are the prerequisites for electronic warfare to analyze and identify information and judge the level of danger.Signal sorting and identification are based on the received parameter information of multiple interlaced radiation sources,analyze the law,and group the pulses according to the radiation source.After the grouping is completed,further identify,locate and interfere with a single radiation source.There are many types of radar radiation sources,rich systems,and changeable working modes.New difficulties arise in radar signal sorting methods.This thesis first studies the traditional single-station signal pre-sorting and PRI main sorting methods.In view of the complex electromagnetic environment,the traditional single-site sorting method is not very robust,and it is easy to increase batches and miss batches.This thesis studies the signal sorting method based on two stations and three stations,analyzes its shortcomings,and proposes a distributed multi-station signal sorting method based on four stations.Then,for the sorted single radiation source,a PRI type recognition algorithm based on a long and short-term memory network(LSTM)network is proposed.The main work of this thesis is as follows:(1)First,this thesis studies the signal pre-sorting of single station based on pulse multiparameter,and simulate the FART-based RF,PW,DOA multi-parameter sorting method.The radar is initially diluted,but some pulse sorting is inaccurate.Then this thesis studies and simulates the main sorting algorithm of PRI signal based on the time of arrival(TOA)and simulate: CDIF,SDIF,PRI conversion method.(2)Aiming at the complex electromagnetic environment,this thesis breaks the conventional single-station sorting idea,improves the two-station and three-station based on pulse arrival time difference(TDOA),and proposes a four-station distributed radar signal sorting method.First,according to the same radiator pulse arriving at different receiving stations with the same TDOA,this thesis studies the pulse grouping method based on dual-station TDOA.The results show that dual-station sorting can initially realize pulse grouping for interlaced and aliased pulse streams.Secondly,this thesis improves the three-station sorting method,extracts the two sets of time difference sets with the same master station's arrival pulse number greater than zero in the time difference sequence composed of the master station and the two slave stations,constructs the TDOA correlation matrix,and analyzes the correlation.Experiments show that this method can effectively eliminate some false time difference pairs.Finally,this thesis proposes a method based on the four-station TDOA group,according to the least square solution of the estimated position,to calculate the residual error,and further eliminates the false TDOA group.And a reasonable threshold can be set to eliminate false TDOA groups.The simulation results show that this method can eliminate the false TDOA group of re-frequency aliased signals.(3)This thesis also studies the single radiation source pulses that have been sorted,and proposes PRI type recognition based on the LSTM network for the three situations of conventional signals,missing pulse signals and interference pulse signals.First of all,this thesis constructs conventional,jitter,jitter,up-slip,and down-slide PRI data sets.Second,we build an environment based on the pytorch deep learning framework in the Linux system,and build an LSTM network to automatically extract features.At the same time,we observe the loss function and accuracy rate,training set and validation set changes with the training cycle and draw real-time graphs.Finally,we evaluate the recognition effect of the LSTM network based on the accuracy rate,precision rate,recall rate,F1 score and other indicators.Theory and research show that PRI type recognition has high accuracy and is insensitive to missing pulses and interference pulses,which is of practical significance.