Research On Key Technologies Of Multi-Function Radar Working Mode Recognition

Since the 1960 s,the phased array radar has become more mature and has been widely used.Compared with the traditional radar antenna,the phased array radar has the ability of beam direction,beam shape,and the ability to form multi-wave formation.Multi-Function Radar(MFR)appears in this context,it is a large-scale dynamic system that combines the powerful data processing of computer and the beam control capacity of the phased array radar.Because MFR has various abilities such as search,interception,tracking,distance detection,and guidance,it is widely deployed and applied.The behavior intention of judging the target MFR based on the intercepted signal is the technical difficulty of solving the in the field of electronic reconnaissance.This article uses the sorted single MFR pulse signal as the research object,and research on the MFR working mode recognition technology.The main work is as follows:1.The unsupervised extraction method of the MFR waveform unit was studied.First of all,based on the waveform unit parameter model held by AN/SPY-1,the MFR signal waveform library is constructed,which can generate the MFR original signal according to the requirements to follow the research of MFR behavior recognition.Secondly,starting with the characteristics of the original MFR signal generated,an MFR waveform extraction method of Step-By-Step Variable Threshold Isolaright Forests(SVT-IForests)is proposed.The method of mentioning makes full use of the multi-parameters of the MFR signal joint change law,advance the detection of leakage pulses and pseudo pulses,and then use SVT-IForests to process pulse parameters,search for the starting pulse of the waveform unit.This method does not require prior knowledge such as MFR waveform libraries,and is suitable for analysis of non-cooperative MFR.Experiments show that this method has greatly reduced the computing complexity while ensuring the accuracy and robustness,which lays the foundation for the subsequent classification and model construction of the waveform unit.2.The method of classification and model reconstruction of non-cooperative MFR waveforms is studied.This article starts from the MFR-level structure,and uses non-cooperative waveform units as the input data.It proposes a method of classification and model reconstruction of waveform units.Under the constraints of Match Level(ML),this method perform multiparameters matching of the waveform units to obtain the similarity between the waveform units;then use the recursive expansion operation to complete the classification of the waveform units with the waveform units pairing score matrix;further the matching state matrix based on similar waveform units can rebuild the waveform unit model.The typical experimental results of the settings show that the method mentioned can complete the classification of the waveform unit under small data volume and significant noise conditions,and can reconstruct the waveform model with a small error.This method can reduce the MFR signal parameters into radar words in discrete domain,which provides analysis objects for subsequent radar phrases mining and working mode recognition.3.Research the unsupervised extraction of radar phrases and working mode recognition method of non-cooperative MFR.According to the MFR sentence model,this article consists of the radar signal composed of radar sequences as sentences that conform to a certain grammar law,and propose an MFR radar phrase mining and pattern recognition method based on radar words sequence.Based on the theory of coding,this article establishes a semantic encoding model of the radar sequence.The model consists of two parts: coding dictionary and coding sequence.Subsequently,this article uses an iterative update dictionary and encoding sequence method to optimize the entire coding model to ensure that the encoding dictionary can accurately describe the radar words sequence;in addition,in order to overcome the adverse effects of the lack of some radar words,this article is merged according to the variables of the complexity of the encoding model to ensure the simplicity of the encoding dictionary;finally,in order to improve the accuracy of working mode recognition,this article integrates the expansion phrase based on frequent collaterals,which can help complete the identification of the error radar phrase.This article verifies the feasibility and performance of the method mentioned based on the two MFR working mode and the radar phrase table of AN/SPY-1 and Mercury.The experimental results show that the method can achieve the expected purpose under the condition of small data volume and radar deficiency,and the experiments on the two MFR data sets show the performance advantages of the comparison method.