The Study Of Radioactive Nuclide Identification Method Based On Fuzzy Decision Tree

The wide application of nuclear technology in various fields has strongly promoted the development of national production and economy.Since the dangerousness and the specificity of the radiative materials,the real-time surveillance during their production process and storage is strictly necessary.The measurement and analysis of gamma photons emitted by radioactive sources is one of the reliable methods for the detection and identification of radionuclide.However,due to the effects of environmental background radiation and shields,the pulse signals detected by the detector are extremely weak.In addition,the low-energy characteristic peak may be affected due to the energy resolution of the detector.It is difficult to extract features effectively using common time-frequency analysis methods,so the detection and identification of radioactive materials is hard to achieve.The gamma-ray spectrums of radioactive nuclear radiation source are taken as research objects in this paper,which aims to explore their feature extraction methods and solves the problem of radionuclide identification.The main contents of this paper include the following aspects:Based on the theory of gamma spectroscopy production,this paper innovates the feature extraction method of gamma-ray spectrums and introduces the idea of sparse decomposition into the feature extraction of gamma spectrums.The traditional sparse dictionary construction method is improved.The peak information which has difficulty to be found in low-resolution spectrums is converted to sparse decomposition coefficient with three kinds of different over-complete dictionary,which enhances the degree of differentiation.The construction method of fuzzy decision tree is improved for the characteristics of gamma-ray spectrums,which effectively exerts the advantages of fuzzy decision tree model and avoids giving the membership values relying on expert knowledge.The recognition results of single and mixed nuclides show that this method can recognize low-resolution spectrums effectively.This paper studies nuclide spectrums feature extraction and recognition algorithm based on the fuzzy decision tree algorithm and sparse decomposition.Under the background of complex noise and low detection rate,the algorithm achieves good recognition effect for ⁶⁰Co,¹³⁷Cs and other nuclide energy spectrums,the highest recognition accuracy can reach 96.39%.