Research On Recognition Method Of Atmospheric Nuclear Explosion Optical Radiation Under Optical Interference Conditions

The ability to detect and identify land-based atmospheric nuclear explosions is very important for our country’s rapid response capability to deal with nuclear threats.The optical radiation effect of atmospheric nuclear explosions has obvious characteristics,and the relevant theories are mature,which are suitable for nuclear explosion identification,explosion method identification and yield estimation.This thesis focuses on the identification of optical radiation signals of atmospheric nuclear explosions by land-based nuclear explosion monitoring equipment under optical interference conditions.The optical interference signals under conditions such as lightning,sunlight background,light reflected by rainy weather and water surface are collected through experiments.Using modern digital signal processing and pattern recognition methods,the target signal and optical interference signals are analyzed,and the identification method of nuclear explosion optical radiation signals under optical interference conditions is explored,which provides ideas and solutions for effectively reducing the false alarm rate and missed alarm rate in the detection of nuclear explosion optical radiation signals and improving the existing detection and recognition methods.The main contents of this article are as follows:（1）The research status of atmospheric nuclear explosion optical radiation signals identification is reviewed,and several existing methods are analyzed and summarized,which are the conditional matching method based on discrete criterion,the identification method based on signal time-domain optical bimodal characteristic,and the signal frequency-domain characteristic-based method.The main problems of these methods are that the generalization ability is not strong due to the small sample size,and there is little research on the optical interference signals that cause the increase of the false alarm rate and missed alarm rate.（2）Study the generation mechanism and propagation characteristics of atmospheric nuclear explosion optical radiation signals,summarize their physical characteristics,provide a basis for signal feature extraction and selection.On the basis of fully studying its characteristics,the simulation and evaluation of the optical radiation signals of the atmospheric nuclear explosion will be carried out,which will lay the foundation for obtaining a sufficient target signal sample size to meet the experimental needs in the follow-up research.（3）Study the acquisition and preprocessing of the target signal and the optical interference signals.In the signal acquisition part,a total of 18,798 signals of various types were collected.Among them,the number of target signal samples was effectively expanded from 9 to 1,009 through the historical data of the actual nuclear explosion and signal simulation methods;by building a portable optical signal acquisition device,a total of 9,789 optical interference signals of various types were collected and 8,000optical radiation signals of various types of noise-added target signals were synthesized according to different intensity ratios,effectively making up for the research on optical interference signals.In the signal preprocessing part,DC elimination,shape-preserving segmental interpolation,digital Butterworth low-pass filter,five-point weighted smoothing filter,wavelet threshold method denoising based on general threshold criterion and normalization method are used respectively.Finally,a data set consisting of target signals,optical interference signals and noise-added target signals are obtained,which provides a standard and unified data source for subsequent experimental research.（4）Study the extraction and selection of signal data features.By taking the optical bimodal characteristic of target signals and its"fast first and then slow"change rule as the starting point,through the analysis of the time and frequency domain characteristics of target signals and optical interference signals,a variety of time-frequency characteristics are extracted.Then,through the method of sequence optimization and Gamma Test-based feature selection,we finally select the four characteristics of time domain double peaks characteristic,signal pulse rising edge,double peak slope ratio,and the ratio of the second maximum illumination time to the minimum illumination time of the signal T_2max/T_min,each of which has its corresponding basic parameters of the nuclear explosion optical radiation time-domain waveform and the physical process corresponding to the signal generation of the signal,and it can comprehensively reflect the"uniqueness"of the nuclear explosion optical radiation signals,which is the basis for subsequent identification research.（5）To study different identification methods of atmospheric nuclear explosion optical radiation signals.In this thesis,the recognition methods based on artificial neural network,support vector machine and fuzzy theory are studied,and the recognition method based on fuzzy theory is mainly studied and designed.In the recognition of target signals and lightning interference signals by these methods,the recognition method based on fuzzy theory is the best,the recognition rate of the target signal has reached 100%,and the comprehensive recognition rate of the lightning interference signal has reached 99.89%.In the recognition of the noise-added nuclear explosion optical radiation signals,the fuzzy recognition method also has the best recognition effect and better anti-interference ability.Finally,through the experimental verification,the effectiveness of the method is strongly proved,and the comparison with the traditional discrete criterion-based conditional matching method is realized.