Fast Near-Field Frequency-Diverse Computational Imaging Based on End-to-End Deep-Learning Network

Area of Science:

• Electromagnetics and Applied Physics
• Computational Imaging
• Metasurface Antennas

Background:

• Metasurface antennas offer advanced wave sculpting but face trade-offs between bandwidth and radiation pattern characteristics.
• Existing computational imaging methods, like matched filters and sparsity-driven algorithms, struggle with low sampling ratios and high computational complexity.
• A need exists for advanced reconstruction techniques to overcome limitations in current metasurface antenna imaging.

Purpose of the Study:

• To develop a novel computational imaging approach using deep convolutional neural networks (CNNs) for metasurface antennas.
• To address the challenges of low scene sampling ratios and high computational complexity in current imaging techniques.
• To improve the reconstruction accuracy and efficiency for various targets, from point-size objects to complex scenes.

Main Methods:

• Integration of deep convolutional neural network (CNN) principles with computational imaging.
• Development of a trained reconstruction network capable of handling high correlation of measurement modes and low scene sampling ratios.
• Comparison against traditional matched filter and compressed sensing reconstruction techniques.

Main Results:

• The proposed CNN-based method effectively reconstructs both point-size objects and complex targets with high accuracy and speed.
• Demonstrated ability to handle high correlation of measurement modes and low scene sampling ratios, outperforming existing methods.
• Significant mitigation of computational burden and requirements for large operation frequency bands.

Conclusions:

• The deep convolutional neural network approach offers a powerful solution for computational imaging with metasurface antennas.
• This method overcomes key limitations of traditional algorithms, enabling more efficient and accurate image reconstruction.
• The findings pave the way for enhanced performance and broader applicability of metasurface antenna systems.