物理
光学
涡流
旋涡
图像处理
散射
物理光学
光散射
折射率
光纤
光学成像
材料科学
领域(数学)
信号处理
作者
Xuan Zhang,JiaHui Chen,Jiahao Zhao,Yue Li,Haowei Wang,Yongnan Li,Chenghou Tu,Hui‐Tian Wang
出处
期刊:Photonics Research
[Optica Publishing Group]
日期:2026-01-20
卷期号:14 (4): 1616-1616
被引量:1
摘要
Vortex beams have achieved remarkable success in optical communications, theoretically enabling unlimited increases in transmission capacity. Therefore, it is crucial to accurately recognize the vortex mode in different scattering environments. Here, we propose a method that combines the speckle-correlation theory with the convolutional neural network (CNN) to effectively and precisely recognize the mode information of superimposed vortex beams under different scattering media. The principle is that each cross-correlation pattern of two related speckle images formed by the in-phase superimposed and out-of-phase superimposed vortex beams exhibits a unique intensity distribution characteristic, which remains invariant across different scattering media. We use a CNN to recognize these intuitive intensity distributions, thereby obtaining the modes of the incident superimposed vortex beams. We use cross-correlation patterns generated by ground glass as the training set, and those generated by fresh egg membranes and chicken breast tissue as the test set. With our method, an accuracy of 95.8% is reached after only 13 training epochs. For comparison, when we use the raw speckle images directly as the CNN data set, the accuracy drops to only 73.9%, showing the advantage of using cross-correlation patterns as the data set for achieving high accuracy across different scattering media. This work provides a new approach for the transmission of vortex beams in scattering environments, with potential applications in communication, encryption, and biomedicine.
科研通智能强力驱动
Strongly Powered by AbleSci AI