Investigating a Non-Linear Iterative Solver Ensemble Method for Microwave Imaging

We present a simple ensemble method for microwave imaging that combines four different non-linear iterative inversion solutions to the same imaging problem. The goals of this approach are (1) to improve the overall image quality and (2) potentially quantify confidence in the final microwave image of unknown targets. Results obtained with experimental data from a 2D near-field imaging system demonstrate that while overall image quality may be improved, and some regions of the variance between algorithms shows accurate areas of error/uncertainty, each imaging algorithm can fail in similar ways as they are applied to the same target data. This suggests that future work to produce ensemble results from different data (e.g., different frequencies or different transceiver configurations) could be used to improve microwave imaging results and provide estimates of error in unknown targets.