Peak-searching method for low count rate γ spectrum under short-time measurement based on a generative adversarial network

In scenarios such as vehicle radioactivity monitoring and unmanned aerial vehicle radioactivity monitoring, the count rate of the γ spectrum detected by the NaI(Tl) detector is low, the characteristic peak is weak, and the statistical fluctuation is large. When such a γ spectrum is processed with the conventional peak-searching method, the characteristic peak recognition accuracy is low and the nuclide identification rate is reduced. A peak-searching method based on the generative adversarial network (GAN) is proposed in this study for low count rate and short-time measurement of a single nuclide γ spectrum. Compared with the symmetric zero-area (SZA) method, the characteristic peak recognition accuracy of the GAN method is improved, the occurrence probability of false peaks is reduced, and the number of false peaks is decreased. Furthermore, the peak position offset with different time measurement conditions of the GAN method is stable. And the performance under shielding conditions of the GAN method is also better than that of the SZA method.