Rapid radionuclide identification algorithm based on the discrete cosine transform and BP neural network

Traditional radionuclide identification algorithm based on peak detection cannot recognize radioactive material in a short time. This study proposes a rapid radionuclide identification algorithm based on the discrete cosine transform and error back propagation neural network. Detection rate and accurate radionuclide identification distance were used to evaluate the proposed method. Experimental results show that the extracted feature vector of the spectrum is not influenced by time, activity, and distance. The proposed algorithm obtained better results in a relatively authentic environment, and it has the ability to predict the isotopic compositions of the mixed spectrum. The proposed method has a better identification performance for the spectrum of radionuclide masked by shielding material except the gamma rays emitted by related radionuclide are significantly shielded. It is also particularly recommended for the fast radionuclide identification of spectroscopic radiation portal monitors, radioisotope identification devices, and other radiation monitoring instruments.