Radar Radiation Source Intelligent Sorting And Identification Technology

The strength and weakness of the radiation source sorting capability has become the core factor of whether the electronic reconnaissance system can function in the existing electromagnetic environment.In this paper,the intelligent sorting identification of radar radiation sources is the theme,and the factors affecting the sorting identification of radar radiation sources are considered comprehensively to carry out an in-depth study,and the specific research contents of the paper are as follows.(1)To study the integrated learning based radiation source binning technique for the problem of radar radiation source pulse mixing.Integrated learning methods such as random forest and XGboost are applied to the binning of measured radar radiation source data.The experimental results show that the integrated learning algorithm is more sensitive to the selection of features,and the dimensions with smaller correlation are selected as feature parameters,which can get better binning results than the nonintegrated learning model,and the accuracy of the random forest model can reach.95.6%.(2)To address the problem of missing radar radiation source pulses or little pulse information,the B-SMOTE-based category imbalance radar radiation source binning identification technique is studied.The B-SMOTE model is applied to a few categories in the sample for sample expansion before binning.The experimental results show that the samples with different imbalance degrees after B-SMOTE expansion can achieve good sorting results under the supervised learning algorithm,and the accuracy rate can reach up to 95.1%.(3)To address the problems of low binning recognition rate and poor robustness of traditional radar radiation source signal binning recognition methods under low signal-to-noise ratio,the CDBN network-based radiation source signal binning recognition algorithm is studied.The CDBN network and Softmax classifier are used to classify and identify six common radar radiation source simulation signals.The binning results of different signals with different signal-to-noise ratios verify the feasibility of the algorithm.By comparing with the existing sorting recognition algorithms that also use time-frequency images,it is proved that the CDBN network algorithm has better sorting recognition effect and robustness,and the overall sorting recognition rate of 92.41%can be achieved even at a low signal-to-noise ratio of-8dB.