Electromagnetic Near-Field Scanning with a Spatially Sparse Sampling Strategy Utilizing Kriging-DMD

This paper proposes a hybrid method for time-resolved electromagnetic near-field scanning, merging model-based (Gaussian processes regression model, a.k.a. Kriging method) and data-driven (dynamic mode decomposition) techniques. Specifically, Latin hypercube sampling enables spatially sparse measurements, followed by dynamic mode decomposition to analyze resulting sparse spatial-temporal data, extracting frequency information and sparse dynamic modes. The Kriging method is then employed for full-state reconstruction. The proposed approach is evaluated using crossed dipole antennas. Results indicate that, even with a spatial subsampling factor of 130, achieving a fully reconstructed field distribution suitable for engineering applications with frequency information extraction is feasible. This hybrid framework presents a promising avenue to enhance efficiency in electromagnetic near-field measurements, potentially finding applications across diverse electromagnetic measurement scenarios.