Radar Based On Convolutional Ceural Network And Denoising Autoencoder Jamming Effect Evaluation

Jamming effect evaluation as an important part of cognitive electronic warfare,is the key to improve the level of their own operations on the jamming side and radar side.Along with the emergence of intelligent countermeasure equipment,higher requirements have been put forward for assessment intelligence.Traditional evaluation methods are more influenced by human beings and too subjective to adapt to the modern battlefield,and it is especially important to apply artificial intelligence to jamming effect evaluation.Firstly,the radar reconnaissance and jamming technology is sorted out,and the signal parameters that can be detected by radar are summarized through the analysis of radar reconnaissance theory,and the simulation analysis of suppression jamming style and deception jamming style is carried out to lay the foundation for the next evaluation index selection.Next,the theory of radar jamming effect evaluation is introduced,and detailed analysis of jamming effect evaluation criteria and jamming effect evaluation warfare techniques indicators is carried out.The selection of indicators is mainly carried out under two conditions: suppressed interference and spoofed interference.The carrier frequency volatility,peak power variation,re-frequency similarity and average pulse width variation are used as the main evaluation indicators under suppressed interference conditions;the average bandwidth increment and beam dwell time increment are used as the main evaluation indicators under spoofed interference conditions.Secondly,under the conditions of suppression jamming and deception jamming,the jamming effect evaluation model based on convolutional neural network is established.During the training process,observe the change of loss function and accuracy with the number of training times,and draw a real-time curve.The accuracy of convolution neural network is 94% and 90% respectively for the two interference conditions.In order to better analyze the interference effect,a regression prediction model for interference effect evaluation based on convolutional neural network is proposed.The regression coefficients of the test set reach more than 98% under the two interference conditions.The simulation results show that the convolutional neural network can be well used in the evaluation of interference effect.Finally,the jamming effect evaluation method under non-complete conditions is investigated,and a denoising autoencoder is built to fill in the missing indicators and predict the jamming effect evaluation.Fixed single indicator loss experiments,unfixed single indicator loss experiments,fixed double indicator loss experiments and loss experiments with different number of samples are conducted under two interference conditions.The simulation experiments show that the algorithm can effectively fill in the missing indicators and predict the evaluation results.In this thesis,we use convolutional neural network and denoising autoencoder to evaluate the radar interference effect,and verify the feasibility of convolutional neural network to evaluate the interference effect,and use denoising autoencoder to effectively fill and effectively predict the missing radar information under the non-complete condition,which has certain engineering reference significance.