An Equivalent Currents Approach to Generating Simulated Training Data for RF Image Classifiers

Robust classification of RF images via machine learning approaches typically requires significant amounts of training data. Training with synthetic data is of interest in order to achieve sufficient samples for arbitrary scenarios of interest. However, generating high-fidelity target models for use in simulation can require significant manual effort and time. To mitigate this challenge, we present a new strategy for generating target scattering models directly from RF measurements based on the principle of equivalent currents.