Research On SAR Jamming Effect Evaluation Based On Auto Encoder

With the wide application of Synthetic Aperture Radar(SAR)systems in modern high-tech local warfare,SAR countermeasure technology has also gained more and more extensive attention.Among them,SAR jamming effect evaluation,as a key part of SAR countermeasures,is an important basis for evaluating the combat effectiveness of friendly and enemy.Reasonable jamming effect evaluation can provide strategic support and battle situation feedback for accurate and efficient military operations.However,with the continuous development of SAR jamming and anti-jamming technologies,the evaluation of SAR jamming effect is faced with the challenges of numerous evaluation indexes,single evaluation dimension and high computational complexity.At the same time,with the rapid development of machine learning,especially deep learning in the field of radar countermeasures in recent years,concepts such as cognitive radar and cognitive jamming have been proposed one after another.The promotion of intelligent and dynamic closed-loop processing of SAR countermeasure technology has become a technical bottleneck to be broken urgently in the field of SAR countermeasures.Therefore,this paper combines deep learning theory with the problem of SAR jamming effect evaluation,and uses the advantages of deep learning in feature extraction and data mining to study the SAR jamming effect evaluation method based on autoencoder,which is of great significance to improve the performance of jamming effect evaluation.Because the existing SAR jamming effect evaluation mostly draws on the traditional narrowband radar jamming evaluation method,it needs to rely on a large number of index calculations,which has a large amount of calculation,and the evaluation accuracy is greatly affected by the selection of the indexes.Therefore,how to make full use of the image domain features of SAR,combined with the strong feature extraction capability of deep learning theory,to achieve intelligent and precise evaluation of SAR jamming effect is a key issue of current research.In response to the above problems,this paper,under the funding of the National Natural Science Foundation of China and the horizontal project of the Science and Technology Commission,combined with the practical application problems in the evaluation of SAR jamming effect,carried out a research on the evaluation method of SAR jamming effect based on autoencoder.The research focuses on the SAR jamming effect evaluation method based on Convolution Autoencoder(CAE)and the SAR jamming effect evaluation method based on Variational Autoencoder(VAE).The main research contents are summarized as follows:1.The working principle of the existing SAR jamming effect evaluation method is studied.Firstly,the workflow of the subjective evaluation method of SAR jamming effect is discussed.Secondly,from the signal domain and the image domain,the calculation and physical meaning of the indexes in the objective evaluation method of SAR jamming effect are analyzed in depth.Then,due to the different evaluation angles of different indexes the Back Propagation(BP)neural network is introduced into the jamming effect evaluation,which solves the problems of single evaluation index and low evaluation accuracy in the traditional evaluation method.Finally,the effectiveness of the method in SAR jamming effect evaluation is verified through simulation experiments,which lays a theoretical foundation for the subsequent SAR jamming effect evaluation methods based on the autoencoder structure.2.Aiming at the problems that the existing SAR jamming effect evaluation methods have many calculation steps and a large amount of calculation,a CAE-based SAR jamming effect evaluation method is proposed.Firstly,the method uses the dual processing characteristics of CAE encoding and decoding to extract feature matrix that can accurately characterize the input SAR image.Then,the feature matrix is used as the input of the SAR jamming effect evaluation network,and the adaptive and accurate evaluation of the SAR jamming effect is carried out.In the network training process,two loss functions,the evaluation accuracy loss and the image reconstruction loss,are used together to improve the evaluation accuracy while improving the accuracy of feature extraction.Finally,the experimental simulation can prove that the method has a good evaluation effect on SAR jamming images.The evaluation accuracy can reach 96.8% on the test set.3.Aiming at the problems that the existing SAR jamming effect evaluation methods lack the ability to extract features,which makes the evaluation accuracy cannot meet the real-time countermeasure requirements and the network convergence speed is slow,a VAE-based SAR jamming effect evaluation method is proposed.Firstly,the method utilizes convolution and pooling operations to perform preliminary feature extraction on the SAR images to be evaluated.Secondly,the extracted preliminary features are used as the input of the VAE module to find latent features in its data distribution.The hidden features encoded by VAE are used as the input of the classifier,and the final evaluation results are obtained through softmax regression processing.In order to ensure the reliability of the extracted features,the network optimizes the iterative network by combining the three loss functions of the evaluation accuracy loss,the image reconstruction loss and the KL divergence of the VAE module.Finally,compared with the CAE-based jamming effect evaluation method,the results show that the VAE-based SAR jamming effect evaluation method has converged faster and the evaluation accuracy is higher.The evaluation accuracy can reach 98.8% on the test set.