Research On Target RCS Prediction Method Of Cognitive Radar

Cognitive radar is a new type of radar that can interact and sense with the working environment,and then actively adjust its own parameters.Radar Cross Section(RCS)is a physical quantity that measures the strength of the target scattered electromagnetic waves,which affects the radar echo quality to a certain extent.As far as the working mechanism of the radar is concerned,a large amount of target information is carried in the echo signal,and radar signal processing is to extract useful information of the target from the echo.In radar resource allocation,if the RCS value of the target can be predicted in advance,provide RCS a priori information to adjust the radar resource allocation scheme to ensure the quality of the target scattered echo;In target detection,by predicting the target RCS and appropriately adjusting the detection threshold,the missed detection rate can be reduced;In the target recognition classification,the accurate acquisition and analysis of the target RCS helps to realize the accurate recognition of the radar target.Therefore,the analysis and prediction of the target RCS is of great significance to ensure the efficient work of the radar.Based on this,this paper studies the target RCS prediction method based on time series,the main work is as follows:Study the target RCS estimation method based on echo;At present,traditional detection methods include matched filtering,MTI,MTD,etc.During the coherent accumulation of high-speed moving targets,MTI is prone to problems such as reducing echo energy and moving across units,resulting in loss of echo energy and reducing the calculation accuracy of the target RCS.In this regard,signal processing methods such as speed search compensation,echo interpolation,distance migration correction,CLEAN,etc.are studied to reduce the loss of echo energy,improve the estimation accuracy of the target RCS,and provide better data for the analysis and prediction of the target RCS Guarantee.At the same time,the method of fitting the target RCS under the narrowband signal based on the broadband demodulated frequency data is studied,and the measured data is processed to provide data verification for the subsequent prediction method.Combining the characteristics of timing fluctuations and timing correlation of target RCS,the basic theories of autoregressive model and Markov model are studied;In the autoregressive model,the RCS time series value near the historical moment is used as a prediction reference,and the order of the model is determined and the weight parameters are optimized through a large amount of data to realize the prediction of the autoregressive model;In the Markov model,it is possible to adaptively adjust the transition probability between models,model selection probability,historical reference probability,etc.based on historical statistics.By updating the RCS sequence,model probability,etc.,the model expression tends to be real changes,and Markov is realized.prediction.Researched the prediction method under the deep learning framework;First,the introduction of recurrent neural network(RNN),through the RNN network long-term historical memory ability,a large number of network parameters and network optimization algorithms,from the historical sequence to fit the target RCS time series changes to achieve the target RCS prediction.Finally,the introduction of long-term and short-term memory networks(LSTM),compared with RNN networks,LSTM network structure of input,output gates and memory units and other computing units can achieve selective reference to input data,through network optimization can reduce far deviations Invalid data interference of the change trend,so as to achieve accurate trend extraction of RCS changes,improve network model reliability,and improve target RCS prediction performance.