Source：npj Nanophotonics
Authors：Yuzhou Song、Yifei Zhang、Xiaoyuan Liu、Takuo Tanaka、Mu Ku Chen、Zihan Geng
Published：2025-07-01
DOI：10.1038/s44226-025-00045-9
Link：https://www.nature.com/articles/s44310-025-00070-9
Core Research Progress
Traditional 3D imaging based on meta-lenses suffers from limited accuracy in scenarios with weak textures and missing features. This study combines binocular meta-lenses and an Optical Clue Fusion Network. Integrating physical stereo depth and machine learning-based depth algorithms, it compensates for imaging defects via adaptive confidence optimization, achieving ultra-high-precision 3D detection with an error below 1% and delivering reliable performance on various complex imaging surfaces.
Technical Application Value
This technology addresses the issues of low imaging accuracy and poor scene adaptability of metalenses. It integrates the merits of light weight, high precision and strong robustness, and greatly improves the reliability of 3D perception. It can be applied to autonomous driving perception, industrial precision inspection, medical imaging diagnosis and other fields, facilitating the industrial implementation of miniaturized and high-precision 3D imaging devices.
Disclaimer
This article is a reorganized and repost of cutting-edge academic information in the industry, only for industry technical exchanges, popular science and industrial dynamic reference, and does not constitute any commercial promotion, technical implementation guidance or investment advice.
All research content, technical achievements and viewpoints of this article are derived from the official published journal papers. The copyright belongs to the original authors and the journal 《npj Nanophotonics》.
If relevant right holders have copyright objections to the content of this article, please contact us. We will verify and revise or delete the content in compliance with regulations in a timely manner.