Machine-learning reprogrammable metasurface imager

TLDR

Conventional microwave imagers require time‑consuming data acquisition or complex reconstruction algorithms, limiting their effectiveness for complex in‑situ sensing and monitoring. We present a real‑time digital‑metasurface imager that can be trained in‑situ to generate radiation patterns optimized by machine learning for measurement modes. The imager is electronically reprogrammed in real time to retrieve the optimized solution for an entire dataset, enabling storage and transfer of full‑resolution raw data in dynamically varying scenes. The system demonstrates high‑accuracy image coding and recognition of handwritten digits and through‑wall body gestures in real time, and opens avenues for intelligent surveillance, rapid data acquisition, and multi‑frequency imaging.

Abstract

Abstract Conventional microwave imagers usually require either time-consuming data acquisition, or complicated reconstruction algorithms for data post-processing, making them largely ineffective for complex in-situ sensing and monitoring. Here, we experimentally report a real-time digital-metasurface imager that can be trained in-situ to generate the radiation patterns required by machine-learning optimized measurement modes. This imager is electronically reprogrammed in real time to access the optimized solution for an entire data set, realizing storage and transfer of full-resolution raw data in dynamically varying scenes. High-accuracy image coding and recognition are demonstrated in situ for various image sets, including hand-written digits and through-wall body gestures, using a single physical hardware imager, reprogrammed in real time. Our electronically controlled metasurface imager opens new venues for intelligent surveillance, fast data acquisition and processing, imaging at various frequencies, and beyond.