Vectorial Digitelligent Optics for High-Resolution Non-Line-of-Sight Imaging

Abstract

Object imaging beyond the direct line of sight is significant for applications in robotic vision, remote sensing, autonomous driving, and many other areas. Reconstruction of a non-line-of-sight (NLOS) screen is a complex inverse problem that comes with ultrafast time-resolved imager requirements and substantial computational demands to extract information from the multi-bounce scattered light. Consequently, the echo signal always suffers from serious deterioration in both intensity and shape, leading to limited resolution and image contrast. Here, we propose a concept of vectorial digitelligent optics for high-resolution NLOS imaging to cancel the wall’s scattering and refocus the light onto hidden targets for enhanced echo. In this approach, the polarization and wavefront of the laser spot are intelligently optimized via a feedback algorithm to form a near-perfect focusing pattern through a random scattering wall. By raster scanning the focusing spot across the object’s surface within the optical-memory-effect range of the wall, we obtain nearly diffraction-limited NLOS imaging with an enhanced signal-to-noise ratio. Our experimental results demonstrate a resolution of 0.40 mm at a distance of 0.35 m, reaching the diffraction limit of the system. Furthermore, we demonstrate that the proposed method is feasible for various complex NLOS scenarios. Our methods may open an avenue for active imaging, communication, and laser wireless power transfer.