Radiation source cross-domain identification model training method and radiation source identification method and device

Abstract

This application provides a method for training a cross-domain radiation source identification model, a radiation source identification method, and an apparatus. The method includes: extracting source domain samples and target domain samples with general features; extracting general feature representations from the source domain samples and target domain samples, and performing element-wise decomposition of the general feature representations to obtain robust and non-robust features of the source domain samples and target domain samples; training the model parameters based on the source domain samples, robust and non-robust features, and a loss function to obtain a trained model; and using the trained model to identify the input target domain signal to be identified, thereby identifying the radiation source of the target domain sample. The method provided in this application improves the training efficiency and identification accuracy of the model.

Images