Research And Application Of Radar Operation Mode Recognition Based On Machine Learning

In order to fulfill diversified air threats and mission requirements,the number of conventional radars is gradually decreasing while a large number of new radar systems with multi-function and multi-purpose are deployed.Thereinto,the multi-function phased array radar is a typical representative which adopts electronically scanned and multi-task management systems.As the threat of multi-function radar to air targets varies greatly under different operation modes,using intercepted radar signals to identify its operation mode is the key problem to threat early warning and assessment,and also the basis of electronic jamming strategy selection and jamming resource allocation.Therefore,it is of great military significance and value to identify the radar operation modes accurately.As the traditional parameter matching method is unable to satisfy the requirements of operation mode recognition in complex electromagnetic environments,machine learning methods are applied to operation mode recognition in this thesis.And the methods of radar operation mode recognition based on machine learning under the conditions of no prior information,large sample database and small sample database are studied respectively.The algorithms are verified through simulation experiments.The methods of radar operation mode recognition proposed in this thesis are of great significance to the research of other electronic reconnaissance technology for reference.The main work of this thesis is as follows:In this thesis,the unknown radar operation mode recognition based on clustering analysis is realized without prior information.Using K-means algorithm,ISODATA algorithm,fuzzy C-means algorithm and fuzzy ISODATA algorithm,the simulation experiments of radar operation mode recognition based on pulse description parameters are carried out.The experimental results verify the effectiveness of the clustering analysis algorithms and prove that the fuzzy ISODATA algorithm has higher accuracy than the other three algorithms.In this thesis,the radar operation mode recognition based on neural networks is realized under the condition of prior information.Firstly,BP neural network algorithm is applied to large sample recognition.Secondly,the Siamese neural network algorithm based on convolutional network is proposed and applied to small sample recognition.In this thesis,the simulation experiments of above two algorithms are carried out respectively.The results show that the two algorithms can achieve preferable recognition effect,and the average classification accuracy of both algorithms is above 87%.Furthermore,the accuracy of the two algorithms is higher than that of the clustering analysis algorithm in the corresponding simulation scene,which fully proves that the radar operation mode recognition algorithm based on neural networks can achieve better recognition effect than that of the clustering analysis algorithms.