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.