Methods And Experiments For Weak Nuclear Signal Detection And Nuclide Identification

In the application of contactless radioactive materials, such as Arms Control Inspection, Counter Nuclear Smuggling and Nuclear Emergency Detecting, sensitive detection of nuclear radiation signal and recognition of nuclides are very important.Owing to background radiation and shield configuration, the radioactive pulse signals captured by sensors are extremely weak, indicating that the amplitude and occurrence time of the weak nuclear pulses are not effectively detected by traditional time-domain and frequency-domain methods. Moreover, limited by the energy resolution of detectors and adverse effect of the background noise and Compton scattering, the signal-to-noise ratio of the observed signals is very low so that the Pulse Height Spectrum (PHS) of the radioactive materials cannot be calculated accurately, resulting in the low rate of identification and detection of radioactive materials. Thus, exploring on how to extract weak signals from heavy background noise and identify nuclides is important theoretical value and practical significance as well. In this dissertation, we are exploring the sparse reconstruction of radioactive signals and classification of nuclides, through improving the detecting capability in the strong noise background, to cope with the problem of detecting weak nuclear pulse signals and the y spectral distortion under the complex detecting situation,thereby fulfilling the task of extracting weak nuclear signals and identifying the nuclides. The major contributions of this dissertation are as follows:(1) In the process of nuclear radiation detection, the statistic model of noise(including impulsive noise, shot noise, Rayleigh noise, Gaussian noise, etc.) is constructed to investigate the noise occurrence mechanism and to analyze the influence of background noise on radioactive signals, thus providing the foundation for the model of detecting nuclear pulse signals.(2) A pulse signal detection method based on sparse representation is proposed to extract useful faint nuclear signals in strong background noise. Utilizing the signal-noise mixing model, noise statistical model and Gabor transformation, a redundant dictionary of underlying signals is constructed to restrain typical ambient noises and a corresponding optimization algorithm is developed by sparse reconstructing. Moreover,a parallel detection method is proposed by sliding window detection in order to improve the performance of the algorithm, indicating that this new algorithm is effective to realtimely detect radioactive pulse signals..(3) To identify nuclides, caused by Gamma spectrum difference from the different measuring conditions including environment and time, the transformation of energy spectrum and the transferred model of nuclides identification are explored. On the one hand, the standard energy spectrum data is used to create a standard energy spectrum matrix. A transferring matrix of energy spectrum in current situation is constructed and learned to improve the differences of energy spectrum due to environment. Thereby the nuclides identification based on the model of nuclide identification obtained from a standard energy spectrum can be realized. On the other hand, using sparse representation and Singular Value Decomposition (SVD), a Gamma-spectrum feature extraction method is proposed and a method of nuclide identification based on the support vector machine (SVM) is investigated. Simulation and experimental results verify the feasibility and effectiveness of the proposed methods.(4) To verify the effectiveness of the proposed radioactive signal detection method and nuclide recognition algorithms in actual conditions, we design and develop an experimental platform, which consists of scintillation detector for weak signal pre-amplifier, analog filter, data acquisition hardware. Then the relevant simulation experiments are carried out by employing the proposed signal detection methods and nuclide recognition algorithms in typical detection situations, such as stations, airports,custom passes, radioactive detection in ocean and so on. The experimental results show that the proposed methods can identify the majority of common nuclides.