Neutron/Gamma Discrimination Study Based on Non-Negative Matrix Factorization and Support Vector Machine

This study addresses the limitations and inefficiencies inherent in traditional discrimination models by introducing an innovative neutron-gamma discrimination model that integrates Non-Negative Matrix Factorization (NMF) with Support Vector Machine (SVM). The proposed NMF-SVM model adeptly extracts global features from the gathered data and rigorously learns from the feature matrix, thereby facilitating precise neutron-gamma discrimination within mixed radiation fields. A comprehensive comparative analysis, juxtaposing the NMF-SVM model with existing methods such as the Charge Comparison Method, the Fractal Spectrum Method, and the Back Propagation Neural Network Discrimination Model, is conducted to ascertain its accuracy and robustness in discrimination. The findings reveal that the NMF-SVM model offers substantial improvements over conventional models, particularly in terms of discrimination speed and accuracy, thereby underscoring its potential for widespread application in radiation detection and measurement.