-
随着同步辐射技术的发展和光源相干性的提升, 叠层相干衍射成像(ptychography)得到快速发展. 叠层相干衍射成像算法解决了传统相干衍射成像算法收敛速度较慢、容易陷入局部最优解和算法停滞等问题, 具有成像视场大、算法鲁棒性强、对误差容忍性高、应用范围广等优点, 正成为相干衍射成像领域的热点研究方向. 本文首先介绍了叠层相干衍射成像算法提出的背景; 然后详细总结了叠层相干衍射成像算法的发展脉络、主要的算法流程以及应用场景, 并且介绍了叠层相干衍射成像与人工智能结合的新算法及应用潜力; 最后介绍了叠层相干衍射成像算法具体的并行化实现及常用软件包. 本文有助于建立叠层相干衍射成像领域算法本身、人工智能以及计算方法全局研究视角, 对于促进叠层相干衍射成像方法学的系统发展具有重要的参考意义.With the development of synchrotron radiation technology and the improvement of light source coherence, ptychography has developed rapidly. Ptychography algorithm solves the problems of slow convergence and easily falls into the local optimal solution and stagnation of the traditional coherent diffraction imaging algorithm. It has the advantages of large imaging field of view, robustness of algorithm, high tolerance to error and wide range of applications, and is becoming a hot research direction in the field of coherent diffraction imaging. Ptychography reconstructs the complex amplitude distribution and illumination light of the sample by iterative algorithms, which can theoretically reach the resolution of the diffraction limit. It has excellent applications in the fields of wavefront detection, phase imaging and optical metrology. This paper first introduces the background of the proposed ptychography algorithm and briefly describes the problem of coherent diffraction imaging algorithm and its development, and then summarizes the development of ptychography algorithm in detail, mainly including the mainstream algorithm of ptychography and its kernel. This paper then describes in detail the improvement of algorithms corresponding to the improvement of the efficiency of ptychography experiments, correction of position errors and the effect of illumination light multi-modal, and elaborates the algorithm flow. After analyzing the possible intersection of diffraction imaging and neural networks in the field of artificial intelligence, this paper introduces new algorithms with combining ptychography with artificial intelligence. New algorithms with combining ptychography with neural networks will have new potential applications in generality, accuracy and robustness. Finally, a specific parallelization implementation of the ptychography algorithm and common software packages are presented. The logic for writing the parallelization of the algorithm implementation of each package and the corresponding advantages and disadvantages of the packages are described in detail. The characteristics and performance of each package are then listed for reference. This paper helps to establish a global perspective of the algorithm itself, artificial intelligence and computational methods in the field of ptychography, and presents an important reference for systematically developing the ptychography method.
-
Keywords:
- ptychography/
- algorithm/
- imaging methodology/
- artificial intelligence
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] -
Ptycholib/Ptychopy PyNX SHARP PtychoShelves 集群名 Argonne ALCF ESRF cluster — PSI GPU型号 NVIDIA K80/A100 NVIDIA V100 GTX Titan Nvidia V100 算法 ePIE DM RAAR DM 单次迭代时间/ns 9.3/— 2.5 1.4 1.03 Nvidia A100/ns 2.8/— 0.7 0.07 0.3 
下载:
导出CSV
软件 算法支持 GPU并行支持 GUI界面 编程/语言/编译 开源 标准化重建时间[76]/(ns·ite–1) Ptychopy DM, ePIE, LSQML $ \surd $ $ \surd $ Python $ \surd $ 9.3 PyNX DM, AP, ML $ \surd $ — Python $ \surd $ 2.5 SHARP RAAR&独立编写 $ \surd $ $ \surd $ — — 1.4 Ptypy DM, ePIE, ML $ \surd $ — Python — — PtychoShelves DM, ML $ \surd $ — Matlab $ \surd $ 1.03/0.7 Ptychography4.0 独立编写 $ \surd $ — Python $ \surd $ — 下载:
导出CSV
-
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95]
下载:
计量
- 文章访问数:8078
- PDF下载量:442
- 被引次数:0