基于卷积神经网络的高阶径向涡旋光束叠加态OAM模式识别方法  

Recognizing OAM mode of high-order radial vortex beam superposition states employing convolutional neural network

作　　者：王姣 王辰白 谭振坤 雷思琛 吴鹏飞[3] 王向辉 邓莉君 WANG Jiao;WANG ChenBai;TAN ZhenKun;LEI SiChen;WU PengFei;WANG XiangHui;DENG LiJun(School of Electronic Information and Artificial Intelligence,Shaanxi University of Science and Technology,Xi’an 710021,China;School of Opto-electronical Engineering,Xi’an Technological University,Xi’an 710021,China;School of Automation and Information Engineering,Xi’an University of Technology,Xi’an 710048,China;School of Physics and Electrical Engineering,Weinan Normal University,Weinan 714099,China)

机构地区：[1]陕西科技大学电子信息与人工智能学院,西安710021 [2]西安工业大学光电工程学院,西安710021 [3]西安理工大学自动化与信息工程学院,西安710048 [4]渭南师范学院物理与电气工程学院,渭南714099

出　　处：《中国科学：物理学、力学、天文学》2024年第8期61-68,共8页Scientia Sinica Physica,Mechanica & Astronomica

基　　金：国家自然科学基金(编号:62101313);西安市高校院所科技人员服务企业项目(编号:22GXFW0004,22GXFW0074,22GXFW0050);陕西省2022年重点研发计划工业领域一般项目(编号:2022GY-100);陕西省科协青年人才支持项目(编号:20220142);陕西省自然科学基础研究计划面上项目(编号:2023-JC-YB-484);碑林区2023年应用技术研发储备工程项目(编号:GX2342)资助。

摘　　要：涡旋光束轨道角动量(Orbital Angular Momentum,OAM)模式的识别属于OAM复用通信系统中的关键技术之一.本文以高阶拉盖尔-高斯(Laguerre-Gaussian,LG)光束为研究对象,提出了一种基于卷积神经网络(Convolutional Neural Network,CNN)的高阶涡旋光束叠加态OAM模式识别方法.将经大气湍流传输后的高阶LG光束叠加态强度图像作为基础训练CNN模型,在给定训练数据的情况下,可以准确高效地对不同条件下高阶LG光束叠加态的OAM模式进行分类与识别.重点讨论了不同大气湍流强度、传输距离、光束截面完整程度和模式间隔对OAM模式识别准确率的影响.研究结果表明,CNN能够在强大气湍流环境下以高于99%的准确率对固定的OAM模式集进行分类和识别.本研究成果为大气湍流环境中叠加涡旋光束的智能解码和解复用提供了新思路.In orbital angular momentum(OAM)multiplexing communication systems,one of the primary technologies is the recognition of the OAM modes of vortex beams.High-order Laguerre-Gaussian(LG)beams were the subject of this investigation,and recognition approach of OAM mode high-order vortex beam superposition states based on convolutional neural networks(CNNs)was suggested.The CNNs model was trained using the intensity images of highorder LG beam superposition states propagating through the atmospheric turbulence.Accurate and effective classification and identification of the OAM modes of high-order LG beam superposition states under various situations can be achieved with training data.The present investigation focuses on how different atmospheric turbulence intensities,transmission distances,beam cross-section integrity and mode intervals affect the identification accuracy of OAM mode.Findings from investigations demonstrate that in conditions with serious atmosphere turbulence,CNNs has above 99%accuracy in classifying and recognizing fixed OAM mode sets.The conclusions of the present investigation offer fresh ideas for the intelligent demultiplexing and decoding of vortex beam superposition states in atmospheric turbulence conditions.

关 键 词：无线光通信 卷积神经网络 涡旋光束 轨道角动量 模式识别 

分 类 号：O43[机械工程—光学工程] TP183[理学—光学]