Research On Radar Active Jamming And Its Identification Method

In the process of rapid development of radar and electronic countermeasures,radar active jamming is becoming more and more complex and changeable,more deceptive characteristics,which leads to the survival and combat capability of the radar is threatened.Based on this,in the process of radar and electronic countermeasures,interference suppression plays a very important role.But different jamming signals will have different suppression methods,before taking anti-jamming measures,the interference pattern needs to be identified and classified.Traditional interference recognition requires manual intervention and feature extraction,but due to many interference patterns and difficulty in feature extraction,it is difficult to achieve its versatility and it is difficult to cope with the increasingly complex electronic countermeasures environment.Based on this,this thesis identifies radar active interference based on convolutional neural network and GoogLeNet network and their improvements.In this thesis,the convolutional neural network and GoogLeNet network are used to identify radar active interference,and the GoogLeNet network model is improved to improve the recognition rate of interference signals.The experimental results achieved good results.The specific research content is as follows:Firstly,based on the radar active interference generated by radar chirp signals,this thesis analyzes nine typical radar active jamming and their principles,and establishes a model on this basis to simulate the waveforms in the time,frequency and time-frequency domains of various interference signals,so as to provide theoretical support for the identification of interference signals.Secondly,five time-frequency analysis methods are studied.According to five timefrequency analysis methods,this thesis simulates the time-frequency images of various interference signals,preprocesses them,and obtains a large number of training and test data sets by setting different dry noise ratios(-8d B～10d B)and different parameters.Then,based on CNN,this thesis identifies radar active interference to realize the identification and classification of interference.Taking the dataset generated by multiple synchronous compression transformation as an example,the CNN network model is trained,and the experimental results show that the overall recognition rate of nine radar active interference signals reaches 89.4%.Then,the interference recognition effect of five time-frequency analysis methods is compared and analyzed,and the results show that the interference recognition effect of multiple synchronous compression transformation is the best.Finally,this thesis proposes GoogLeNet-based radar active interference recognition,uses transfer learning,takes the trained GoogLeNet network as the initial value of the radar active interference recognition construction network,and learns the new task of the target domain through network fine-tuning,so as to realize the identification of GoogLeNet-based radar active interference,and the overall recognition rate of interference signals reaches 93.51%.Aiming at the recognition of interference signals,this thesis further improves the recognition rate of radar active interference signals by improving the network model,optimizing the activation function and optimizer.The improved GoogLeNet network achieves a high recognition rate(97.54%)for radar active interference,and under the same training set and training parameters,compared with the unimproved GoogLeNet model,the recognition rate of radar active interference signal is significantly improved.