Aiming at the problems of difficult feature extraction

complex modeling and low recognition rate on complex gamma spectrum under high background and low detection rate that traditional nuclide recognition methods have

a nuclide identification method based on convolution neural network（CNN）is proposed.This method achieves multi-nuclide identification by utilizing CNN to adaptively and implicitly extract image features and classify learning characteristics

which acquires deep level through the training of network model feature extraction layer and classifier.Experiment of conparative analysis is carried out based on Monte Carlo analysis software Geant4 simulation data.Results show that the time complexity of the proposed method is O（n2）.This method avoids the complex process of explicit feature extraction and can realize real-time multi-nuclide identification for some common mononucleotides and mixed nuclides specified by IAEA.