FAExGAN-DA: a GPR data augmentation model based on fuzzy min-max neural network and extended generative adversarial network

Acquiring sufficient data for training machine learning models to identify and detect hidden pipelines using groundpenetrating radar (GPR) imagery is challenging due to the high acquisition costs. We propose FAExGAN-DA, a data augmentation model that utilizes an improved fuzzy min-max neural network and extended generative adversarial network. The model effectively captures higher-order features and potential nonlinear relationships in GPR image data by reconfiguring the generator of the generative adversarial network using a deep convolutional neural network. Moreover, it improves the accuracy of GPR image data classification by enhancing the feature extraction part of the fuzzy min-max neural network. The extended generative adversarial network and improved fuzzy min-max network significantly enhance the quality of generated data and model interpretability. The experimental results on the validation set confirm the superiority of the proposed method for enhancing GPR data. The proposed model was trained using both fake data generated by existing models and data generated by the proposed model in this study for the training of open classifiers.