End-to-end measurement of spatial coherence length using a convolutional neural network
Optics Express · 10.08.2026 · 34 (16) · 29601
DOI: 10.1364/oe.603426 ↗Kurz gefasst
Fachartikel in der Zeitschrift Optics Express, online erschienen am 31.07.2026. Verfasst von Zekun Bi und 3 weiteren Autor:innen. Die Arbeit nennt 23 Literaturangaben. Gefördert von: The Strategic Priority Research Program of the Chinese Academy of Sciences.
Abstract (Originalsprache)
Measuring the spatial coherence length of partially coherent light is fundamentally important for evaluating optical systems and advancing optical engineering. In this paper, we propose an end-to-end deep learning framework capable of retrieving spatial coherence length from a single-shot, fixed slit separation two-dimensional interference pattern. Our approach utilizes a modified ResNet-18 convolutional neural network integrated with a convolutional block attention module to adaptively extract critical coherence features while effectively suppressing background speckle and dark-field noise. To train the network efficiently, a high-fidelity dataset was generated using the Gaussian-Schell model combined with coherent mode decomposition. Experimental validation utilizing a pseudo-thermal light source demonstrated exceptional accuracy, achieving a relative error of less than 2% across a coherence length range of 50 to 300 μm. Furthermore, our study establishes the optimal matching relationship between slit separation and target coherence length required to maximize measurement dynamic range. This compact, high-speed, and robust approach offers a highly effective alternative for real-time optical coherence metrology.
Förderung
- The Strategic Priority Research Program of the Chinese Academy of Sciences
Zitieren
APA
Bi, Z., Ma, Y., Lu, Z., & Wu, X. (2026). End-to-end measurement of spatial coherence length using a convolutional neural network. Optics Express, 34(16), 29601. https://doi.org/10.1364/oe.603426
Vancouver
Bi Z, Ma Y, Lu Z, Wu X. End-to-end measurement of spatial coherence length using a convolutional neural network. Optics Express. 2026;34(16):29601. doi:10.1364/oe.603426
BibTeX
@article{Bi2026,
author = {Bi, Zekun and Ma, Yisi and Lu, Zengxiong and Wu, Xiaobin},
title = {End-to-end measurement of spatial coherence length using a convolutional neural network},
journal = {Optics Express},
year = {2026},
volume = {34},
number = {16},
pages = {29601},
doi = {10.1364/oe.603426}
}
EUVIMED