Metasurface Sensor and Deep Learning Enable Real-Time Hyperspectral Polarimetric Imaging

A research team led by Professor Wei Lu, Professor Xiaoshuang Chen, and Professor Jing Zhou from the Shanghai Institute of Technical Physics, Chinese Academy of Sciences, has developed a high-dimensional optical sensing framework that achieves single-shot hyperspectral full-Stokes real-time imaging. Published in Light: Science & Applications, the framework integrates metasurface physical encoding with artificial intelligence-driven decoding to capture high-dimensional optical light fields.

The proposed system simultaneously reconstructs 167 spectral channels and full-Stokes polarization information across the broadband range of 1150 to 1650 nanometers. In experimental evaluations, the device demonstrated a minimum distinguishable wavelength interval of 3 nanometers, an average polarization reconstruction error as low as 0.14 degrees, and a real-time imaging speed of 55 frames per second.

At the center of the imaging architecture is an infrared meta-sensor that integrates a metasurface with an indium gallium arsenide focal plane array. The system actively utilizes both the polarization multiplexing capability and the structural dispersion characteristics of the metasurface as high-dimensional information encoding degrees of freedom.

Unlike conventional optical designs that attempt to suppress wavelength-dependent dispersion effects, the framework actively harnesses the intrinsic polarization–dispersion coupling of the metasurface. Different polarization states are spatially multiplexed into distinct diffraction channels, while different wavelengths are continuously dispersed along the radial direction through structural dispersion.

To reconstruct the encoded optical information, the researchers combined physical metasurface encoding with a deep neural network designated as HSPI-net. The neural network learns the complex physical mapping relationships of the real optical system, enabling the direct reconstruction of hyperspectral full-Stokes information from a single captured grayscale image.

According to the researchers, utilizing the polarization modulation capability and structural dispersion characteristics of the metasurface simultaneously as high-dimensional optical encoding degrees of freedom allows the system to reconstruct hyperspectral full-Stokes information directly from a single captured frame, facilitating real-time high-dimensional sensing.

The scientists further emphasized that transforming intrinsic polarization–dispersion coupling into a usable information encoding mechanism, rather than treating dispersion as a limitation, provides a new route toward compact, intelligent, and real-time high-dimensional optical sensing systems.

Embedded-platform deployment validation confirmed the system’s operational viability for compact, intelligent sensing hardware. Looking ahead, the researchers forecast that the technique could open new avenues for intelligent optical sensing systems in autonomous driving, infrared target recognition, remote sensing, biomedical imaging, machine vision, and high-dimensional optical communications.

Subject of Research: Real-time hyperspectral and polarimetric light-field imaging

Article Title: New paradigm for real-time high-dimensional hyperspectral–polarimetric light-field imaging

Article References: New paradigm for real-time high-dimensional hyperspectral–polarimetric light-field imaging. Original publication (Bibliographic details unavailable.)

Image Credits: AI Generated

DOI: Not provided

Keywords: hyperspectral imaging, polarimetric imaging, metasurfaces, deep learning, optical sensing, infrared sensors, light-field imaging, HSPI-net, paradigm, real-time, high-dimensional, hyperspectral