Study On Radar Mainlobe Jamming Suppression Method Based On Deep Neural Network

Mainlobe jamming brings severe challenges to radar detection,and effective mainlobe jamming suppression is an urgent problem to be solved.Interrupted sampling repeater jamming based on Digital Radio Frequency Memory(DRFM)technology is a common mainlobe deception jamming style.It makes the jamming signal highly similar to real target signal in time domain,frequency domain and spatial domain by intermittent sampling and forwarding of the intercepted radar signal at low rate.Traditional radar is difficult to identity and suppress the jamming while retaining the target signal,which poses a serious threat to radar target detection and tracking.Based on the above background,this paper focuses on the method of mainlobe jamming suppression and target detection based on deep neural network and the problem of anti-jamming detection under small samples.The main work includes:1.Aiming at the problem of target detection under the background of mainlobe interrupted sampling repeater jamming,this paper models it as an end-to-end integrated processing process from receiving to anti-jamming detection based on deep neural network.The parallel multi-scale convolution kernel,attention mechanism and voice processing system Wave Net network are applied to the design process of detection model,so as to complete the mainlobe jamming suppression and target detection more efficiently.Compared with the existing model,the model designed in this paper achieves better anti-jamming and target detection performance under different radar transmitting waveforms and interrupted sampling repeater jamming styles.2.Aiming at the problem of small samples in radar anti-jamming detection caused by insufficient training samples,a method of mainlobe jamming suppression and target detection based on transfer learning is proposed.In this method,the useful knowledge acquired from the source domain is transferred to the target domain by means of model parameter transfer,that is,large-scale simulation data sets are used to assist in optimizing the detection effect of the model on small sample data.The experimental results based on small sample set show that this method can effectively improve the network training efficiency,and achieve the purpose of obtaining high detection probability with only a few real echo samples.3.Aiming at the performance degradation of anti-jamming detection model under small samples,a method of mainlobe jamming suppression and target detection method based on Wasserstein generative adversarial network is proposed.In order to expand the limited echo data set into a sufficient sample to support training,this paper introduces Wasserstein generative adversarial network,and designs a generation method for radar echo data,which makes the generated samples more diverse while retaining the key structure of the original data.The experimental results show that adding new generated samples to the original training set can alleviate the problem of poor generalization performance caused by insufficient training samples,and effectively improve the anti-jamming performance of the network.Through in-depth exploration and research,this paper further broadens the application scope of deep neural network in radar anti-jamming,and provides the research basis for the future related issues.