Radar Operation Mode Recognition Based On Adaptive Sorting

Radar signal sorting and operation mode recognition are important aspects of electronic reconnaissance,providing important information support for threat level assessment and electronic jamming decisions,which affecting the final trend of electronic warfare.With the rapid increase in the number of radiation sources under complex electromagnetic environments,as well as the lack of a sample library for radar operation modes caused by non-cooperation,it has brought great challenges to electronic reconnaissance.Therefore,this paper carries out the research on signal adaptive sorting and radar operation mode recognition.Firstly,a radar inter-pulse modulation model is constructed to establish the relationship between radar parameters and inter-pulse modulation.Based on the beam scanning method and the characteristics of different operation modes,the temporal characteristics are combed,and the back propagation algorithm is used to explore the performance of operation mode recognition.Next,the adaptive sorting of radar signals is studied.A radar signal adaptive sorting algorithm based on DBSCAN-PRIT-DR is proposed,which employs the K-distance curve to obtain clustering parameters,conducts density-adaptive clustering,and combines pulse repetition interval transformation as well as data rearrangement.Furthermore,due to the absence of radar operating mode sample library,the recognition of radar operation modes under small-sample conditions is studied.The radar parameters are pre-processed,and the conversion from pulse descriptor words to operation mode images is achieved.A convolutional neural network model is constructed to extract local detailed features.Based on the idea of metric learning,a small-sample radar operation mode recognition algorithm based on LDF-KNN is proposed,which combines the K-nearest neighbor algorithm and the intercalation training mechanism.Then,in a low-data scenario,based on the principle of domain adaptation,a pre-trained network model is transferred and a PN-SVM algorithm is proposed for radar operational mode recognition using a support vector machine classifier.Finally,the radar operation mode recognition under large samples is studied,and2D-CNN-GRU based on radar operation mode recognition algorithm with large samples is proposed by using gated recurrent neural unit as well as improving 2D convolutional neural network.Based on this,the radar radiation source operation mode recognition system is realized.Simulation results show that under the condition of 4% pulse loss,the correct rate of signal sorting reaches 87.6%.Under the condition of the same pulse loss rate and 10 samples per class,the correct rate of small-sample operation mode recognition reaches 75.3%.Under the conditions of 100 samples per class,the correct rate of few-sample operation mode recognition reaches 80.1%.Under the conditions of 1000 samples per class,the correct rate of large-sample operation mode recognition reaches 83.5%.This completes the adaptive sorting of radar signals and radar operation mode recognition under different training sample quantities,providing new ideas for intelligent situational awareness on the battlefield.