Three-dimensional (3D) surface imaging technologies, particularly laser-based systems, are crucial for various applications including augmented/virtual reality (AR/VR), autonomous driving, robotics, and mobile face recognition. These systems typically use either time-of-flight (TOF) or structured light (SL) methods for depth estimation. TOF systems measure the time delay between emitted and received laser pulses, while SL systems project structured light patterns onto objects and analyze the distortions to calculate depth. Existing TOF systems often face limitations in FOV and frame rate due to mechanical scanning or limited electronic beam steering capabilities. SL systems, while offering wider FOVs, typically rely on diffractive optical elements (DOEs) or spatial light modulators (SLMs) with limitations in diffraction efficiency and uniformity due to their micron-scale pixel sizes. Metasurfaces, two-dimensional arrays of subwavelength nanostructures, offer a potential solution by enabling precise light modulation with high efficiency at the nanoscale. Previous metasurface-based SL systems have achieved limited FOVs. This research aims to address these limitations by developing a metasurface-based SL system with a significantly enhanced FOV (180°) and high-density dot array projection, enabling improved depth perception for various applications.

The paper reviews existing 3D imaging technologies, focusing on the limitations of TOF and conventional SL methods. TOF systems using mechanical scanning or MEMS suffer from low frame rates and power consumption issues. Electronic beam steering using TCOs and LCs show promise but are limited in FOV. Conventional SL systems using DOEs and SLMs have limitations in diffraction efficiency and uniformity, especially at large angles. While some progress has been made using metasurfaces to improve SL projection, existing approaches are limited to smaller FOVs (120° or less) and a smaller number of diffracted beams. The authors highlight the potential of metasurfaces to overcome these limitations due to their subwavelength pixel pitch, enabling superior FOV and diffraction efficiency compared to conventional approaches.

The proposed metasurface is designed to generate a high-density dot array across a full 180° FOV. The design principle leverages the diffraction effect of individual meta-atoms and the interference effects from their periodic arrangement. The design process involves defining a target spatial frequency distribution, retrieving the phase profile of a single supercell using the Gerchberg-Saxton (GS) algorithm, and then considering the interference effects from the periodic arrangement of supercells using a 2D Dirac comb function. The convolution theorem is applied to determine the final diffraction pattern. Rectangular meta-atoms made of hydrogenated amorphous silicon (a-Si:H) are optimized using rigorous coupled-wave analysis (RCWA) to achieve high polarization conversion efficiency (88%). The metasurface is designed to be polarization-independent, simplifying the optical setup. The intensity uniformity of the diffracted beams is analyzed using both simulations (COMSOL Multiphysics and Rayleigh-Sommerfeld diffraction theory) and experiments. A 1D metasurface is used for experimental validation due to the complexity of measuring the large number of points in a 2D metasurface. The experimental setup utilizes a rotating power meter to measure the intensity of each diffracted order. Depth estimation is demonstrated using a stereo vision system with two cameras. A stereo matching algorithm, utilizing coherent point drift (CPD) for point set registration, is employed to calculate the depth of projected dots on face masks. The nano-PER imprinting method is used to fabricate metasurfaces on curved substrates (safety glasses) for prototype demonstration.

The research successfully demonstrated a metasurface-based SL system capable of projecting approximately 10,000 dots over a 180° FOV. Simulation and experimental results showed good agreement in terms of the number and intensity distribution of diffracted beams, with a root mean square error (RMSE) of 27.48% for a 1D metasurface. The depth estimation experiment using two face masks, one placed on the optical axis and the other at a 50° angle, yielded accurate 3D reconstructions with absolute depth errors of 8.5% and 1.9%, respectively. The nano-PER imprinting method successfully replicated the metasurface onto a curved surface, demonstrating the feasibility of integrating this technology into compact devices like AR glasses. The overall diffraction efficiency of the fabricated metasurface was measured to be 60%, with the zeroth-order efficiency at 32%. The time taken for the stereo matching algorithm was approximately 0.24-0.35 seconds per frame. An additional experiment using a 100 mW laser demonstrated extended operating range.

The results demonstrate the significant improvement in FOV and dot density achieved using the proposed metasurface-based SL system compared to conventional methods. The accurate depth reconstruction of objects at varying angles validates the effectiveness of the design and the stereo matching algorithm. The successful fabrication of metasurfaces on curved surfaces using nano-PER imprinting highlights the scalability and potential for practical applications. The limitations related to power dispersion and resolution at longer distances and larger angles are addressed, proposing solutions such as using higher-power lasers and NIR light sources. The polarization-independent nature of the metasurface simplifies the system design. Overall, the study showcases the potential of metasurface-based SL for creating compact, high-performance 3D imaging systems.

This research successfully demonstrated a metasurface-based structured light system with a significantly enhanced 180° FOV and high-density dot array projection. Accurate depth reconstruction was achieved, and the nano-PER imprinting method enabled scalable fabrication on various substrates. Future work could focus on improving the operating range and resolution by using higher-power lasers, optimizing metasurface designs, and exploring alternative, faster depth estimation algorithms. Integration with on-chip light sources like VCSELs could lead to ultra-compact 3D imaging systems for various applications.

The current system has limited operating range (1m) and resolution at larger distances and viewing angles due to power dispersion and increasing spacing between diffracted beams. The speed of the depth reconstruction algorithm (3-4 fps) is relatively slow compared to real-time applications. The experimental validation focused on a 1D metasurface due to the complexity of measuring many points in a 2D metasurface. The diffraction efficiency of the fabricated metasurface was lower than the theoretical prediction due to fabrication imperfections and coupling between meta-atoms.