Identification Method Of Radar Active Deception Jamming Based On Deep Learning

With the development of electronic jamming technology,particularly the emergence of digital radio frequency memory(DRFM)technology,the electromagnetic jamming environment of modern radars has become increasingly complex.When the radar processes the received signal,it is required to detect the existence of interference and identify its type in time,and take appropriate anti-interference measures to ensure that the radar can correctly detect and track targets.This thesis takes the identification of radar active deception jamming as the core and studies the detection of deception jamming based on DRFM and the identification of small samples.The main contents are summarized as follows:(1)The forwarding deception jamming based on DRFM technology is highly coherent with the real radar echo,which makes it difficult for the radar to distinguish between real and fake targets.To address this issue,the thesis proposes a detection method for DRFM deception jamming based on the Hough transform,which uses short-time fourier transform(STFT)and a two-dimensional constant false alarm rate(CFAR)detector to extract the features of the jamming signal,and the deceptive interference detection is completed by using the Hough transform.This method is based on the characteristics of the interference itself,which results in little dependence on prior information and application scenarios,and low computational complexity.Simulation results demonstrate that the proposed method has strong detection capabilities in low signal-to-noise ratio(SNR)and in various scenarios.(2)Aiming at the problem of the insufficient number of samples in actual radar deception jamming signals,a small-sample-based multimodal radar active deception recognition method is proposed.This method is based on the characteristic parameters extracted from the radar signal and the two modal information of the time-frequency image,trains the multi-modal features using a prototype network,and improves the recognition performance under low SNR with the help of image denoising processing and weighted Euclidean distance.The proposed method achieves radar deception jamming recognition under small sample conditions.Simulation results show that the proposed method exhibits good noise immunity,and the recognition performance is good under the condition of small samples.The test results on simulator data indicate that the method also has good generalization performance.(3)In the process of jamming identification,the available features of towing jamming signals are less,especially when there are few false targets and many real targets,the confusion will be more serious,which will affect the overall recognition performance.To address this problem,the thesis proposes a towing interference recognition method based on Siamese networks.This method reconstructs the data set by considering the towing nature of the interference,uses the Siamese network to identify the towing interference,and performs decision fusion with the multimodal interference identification network to complete the interference identification.The simulation results show that the proposed method can effectively reduce the confusion between towing jammers and other jammers in scenarios with few false targets and multiple targets,and to some degree,improve the overall recognition accuracy of towing jammers.