基于深度学习的粒子场数字全息成像研究进展  

Research progress of particle field digital holography based on deep learning

作　　者：吴羽峰 吴佳琛 郝然 金尚忠 曹良才[3] WU Yufeng;WU Jiachen;HAO Ran;JIN Shangzhong;CAO Liangcai(College of Optical and Electronic Technology,China Jiliang University,Hangzhou 310018,China;Key Laboratory of Zhejiang Province on Modern Measurement Technology and Instruments,Hangzhou 310018,China;State Key Laboratory of Precision Measurement Technology and Instruments,Department of Precision Instrument,Tsinghua University,Beijing 100084,China)

机构地区：[1]中国计量大学光学与电子科技学院,浙江杭州310018 [2]浙江省现代计量测试技术及仪器重点实验室,浙江杭州310018 [3]精密测试技术及仪器国家重点实验室,清华大学精密仪器系,北京100084

出　　处：《应用光学》2020年第4期662-674,共13页Journal of Applied Optics

基　　金：国家自然科学基金(61975182);国家重点研发计划(2018YFF0214904);浙江省重点科技计划项目(2020C03095);之江实验室重点研发项目(2019DE0KF01)。

摘　　要：粒子场的数字全息成像中,由一幅粒子场全息图重建出高精度的三维粒子场分布,是数字全息技术领域的经典问题之一。相比于传统反向重建算法,深度学习算法可以从单个全息图直接重建出三维粒子场来简化算法复杂度,提高计算效率和准确率。介绍国内外研究团队将深度学习算法结合数字全息技术实现粒子场数字全息成像的研究进展,从不同粒子表征方法入手,叙述了支持向量机、全连接神经网络、全卷积网络、U-Net网络、深度神经网络在粒子场数字全息成像中粒子表征及粒子场反向重建过程中的应用原理、实现途径和准确率。最后指出了深度学习算法在这一研究领域的优势及目前基于深度学习算法的不足,并对如何进一步提高该方法的准确率进行了展望。In the digital holography of particle field,reconstructing a high-precision three-dimensional particle field distribution from a particle field hologram is one of the classic problems in the field of digital holography.Compared with the traditional inverse reconstruction algorithm,the deep learning algorithm can directly reconstruct the three-dimensional particle field from a single hologram to simplify the algorithm complexity and improve the calculation efficiency and accuracy rate.The research progress of particle field digital holography in combining deep learning algorithm with digital holography technology by the research teams at home and abroad was introduced.Starting from different methods of particle characterization,the application principles,implementation approach and accuracy rate of the support vector machine,fully connected neural network,fully convolutional network,U-Net network and deep neural network in the process of particle characterization and particle field inverse reconstruction for particle field holography were described.Finally,the advantages and shortcomings of deep learning algorithm in this research field were pointed out,and how to further improve the accuracy of this method was prospected.

关 键 词：粒子场 数字全息 深度学习 神经网络 

分 类 号：TN206[电子电信—物理电子学] O436[机械工程—光学工程]