基于Walsh函数调制的单帧图像深度学习相位反演技术  

Deep learning phase inversion technique for single frame imagebased on Walsh function modulation

作　　者：刘奇 杜应磊[1] 向汝建[1] 李国会[1] 张秋实 向振佼 吴晶[1] 岳献 包安超 游疆 Liu Qi;Du Yinglei;Xiang Rujian;Li Guohui;Zhang Qiushi;Xiang Zhenjiao;Wu Jing;Yue Xian;Bao Anchao;You Jiang(Institute of Applied Electronics,CAEP,Mianyang 621900,China;Graduate School of China Academy of Engineering Physics,Beijing 100088,China)

机构地区：[1]中国工程物理研究院应用电子学研究所,四川绵阳621900 [2]中国工程物理研究院研究生院,北京100088

出　　处：《强激光与粒子束》2024年第6期166-174,共9页High Power Laser and Particle Beams

摘　　要：远场相位反演存在简并态,复原波前容易出现多解问题。相较于传统的迭代算法,结合了相位调制与深度学习的相位反演方法不仅显著降低了计算复杂度,还可有效地解决多解问题。这种方法实时性强,结构简洁,显示出其独特的优势。使用不同的沃尔什函数(Walsh)对相位进行调制,采取深度学习的方法训练卷积神经网络从调制后单帧远场强度图中获得第4~30项Zernike系数从而复原出原始波前,解决了相位反演多解问题。对于3~15 cm大气相干长度的湍流像差的残差波前,其RMS与原始波前RMS的比值可达7.8%。深入研究了Zernike阶数、随机噪声、遮挡以及强度图分辨率等多种因素对波前复原精度的影响。研究结果表明,这种基于深度学习的相位反演方法在复杂的环境中展现出了良好的鲁棒性。The far-field phase inversion exhibits degeneracy states,leading to the problem of encounteringmultiple solutions when recovering the wavefront.In comparison to traditional iterative algorithms,the combination ofphase modulation and deep learning in the phase inversion method not only significantly reduces computationalcomplexity but also effectively solves multi-solution problems.This method possesses strong real-time capabilitiesand a simple structure,showcasing its unique advantages.In this paper,different Walsh functions are used to modulatethe phase,and a deep learning approach is taken to train a convolutional neural network to obtain the 4th-30th orderZernike coefficients from the modulated single-frame far-field intensity maps so as to recover the original wavefront,which solves the problem of multiple solutions of phase inversion.For the residual wavefront of the turbulentaberration of 3-15 cm atmospheric coherence length,the ratio of its RMS to the RMS of the original wavefront canreach 7.8%.In addition,this paper also deeply investigates the effects of various factors such as Zernike order,randomnoise,occlusion,and intensity map resolution on the wavefront recovery accuracy.The results show that this deeplearning-based phase inversion method exhibits good robustness in complex environment.

关 键 词：相位反演 泽尼克系数 相位调制 深度学习 沃尔什函数 

分 类 号：O439[机械工程—光学工程]