Research On Radar Signal Detection And Interference Recognition Technology Based On Deep Learning

Modern warfare has been a five-dimensional warfare of sea,land,air,space and information.Radar as a ‘sentry' plays a vital role in modern warfare.The complex and complicated electromagnetic environment can obscure the eyes of the battlefield-radar,so that the radar in the electronic information war blind,resulting in the loss of war.Modern warfare has evolved into a game of radar technology confrontation between the combatants,so radar technology is particularly important in electronic information warfare,and jamming and anti-jamming is an important research area in radar confrontation.With the rapid development of digital RF storage technology,a smart radar jamming pattern with good autonomous operability,strong flexibility and high similarity to the signal has emerged,which has strong deception and suppression.The traditional anti-interference methods are narrowly targeted,limited,not flexible enough,and poor interference suppression.In the process of rapid development of technology,the artificial intelligence methods such as machine learning and deep learning have developed rapidly.Deep learning can dig the inherent rules of sample data,a large number of researchers have introduced deep learning into the field of radar recognition in recent years,aiming to use intelligent means to achieve effective identification of interference signals,thus solving the problems of traditional anti-jaming methods.In this thesis,a study of radar signal detection and interference identification techniques based on BP neural network,convolutional neural network and recurrent neural network is carried out for the problem of radar anti-jamming.First,to address the problem of existing publicly available radar data sets being scattered and of varying standards,this thesis uses the electronic system-level simulation software System Vue to build a fully autonomous integrated model of radar target echoes for generating radar echo signals.The effectiveness and feasibility of the model are verified by comparing with the measured data.Second,the types of radar dexterous interference based on digital RF storage technology are compared visually,and an integrated model of radar dexterous interference is built independently using the simulation platform.By using the interference machine to sample the measured data and generate the interference signal,the pulse pressure analysis results are consistent with the simulation which confirms the correctness of the model.It provides a new idea to solve the problem of the lack of existing deep learning radar datasets,effectively fills the gap of radar echo datasets,and also provides dataset support for the application of deep learning in the field of radar.Finally,the study of intelligent methods for radar signal detection and interference identification is conducted.Back Propagation Neural Network,Convolutional Neural Networks and Recurrent Neural Network framework models are studied.The model based on RNN for signal double choice detection,jamming double choice and mutilple choice recognition is designed and built independently by Tensorflow.Then compared with models based on BP neural network and CNN,the experimental results show that the models based on RNN and CNN achieve higher recognition rates,which are more suitable for radar signal recognition.