• Chinese Optics Letters
  • Vol. 18, Issue 5, 050602 (2020)
Hongye Li1, Hu Liang4, Qihao Hu1, Meng Wang1、2、3, and Zefeng Wang1、2、3、*
Author Affiliations
  • 1College of Advanced Interdisciplinary Studies, National University of Defense Technology, Changsha 410073, China
  • 2State Key Laboratory of Pulsed Power Laser Technology, Changsha 410073, China
  • 3Hunan Provincial Key Laboratory of High Energy Laser Technology, Changsha 410073, China
  • 4Tianjin Navigation Instruments Research Institute, Tianjin 300131, China
  • show less
    DOI: 10.3788/COL202018.050602 Cite this Article Set citation alerts
    Hongye Li, Hu Liang, Qihao Hu, Meng Wang, Zefeng Wang. Deep learning for position fixing in the micron scale by using convolutional neural networks[J]. Chinese Optics Letters, 2020, 18(5): 050602 Copy Citation Text show less
    Schematic of offset splicing.
    Fig. 1. Schematic of offset splicing.
    Excitation efficiency of each fiber mode versus offset point (offset direction: x axis).
    Fig. 2. Excitation efficiency of each fiber mode versus offset point (offset direction: x axis).
    Auto-correlation function of specklegrams: (a) offset point in the x direction varies from −25 μm to 25 μm; (b) offset point in the whole plane.
    Fig. 3. Auto-correlation function of specklegrams: (a) offset point in the x direction varies from 25μm to 25 μm; (b) offset point in the whole plane.
    Excitation ratio of (a) the third and (b) the seventh lower modes at different offset points.
    Fig. 4. Excitation ratio of (a) the third and (b) the seventh lower modes at different offset points.
    Specklegrams of two centrosymmetric offset points: (a) (6 μm, −5 μm) and (b) (−6 μm, 5 μm).
    Fig. 5. Specklegrams of two centrosymmetric offset points: (a) (6 μm, 5μm) and (b) (6μm, 5 μm).
    (a) Data processing flow and (b) CNN architecture.
    Fig. 6. (a) Data processing flow and (b) CNN architecture.
    Loss versus training epoch (loss, MSE in training set; and val_loss, MSE in validation set).
    Fig. 7. Loss versus training epoch (loss, MSE in training set; and val_loss, MSE in validation set).
    Distribution of difference between the prediction label and validation label.
    Fig. 8. Distribution of difference between the prediction label and validation label.
    Labeled offset points with d higher than 2 μm in the validation set.
    Fig. 9. Labeled offset points with d higher than 2 μm in the validation set.
    Hongye Li, Hu Liang, Qihao Hu, Meng Wang, Zefeng Wang. Deep learning for position fixing in the micron scale by using convolutional neural networks[J]. Chinese Optics Letters, 2020, 18(5): 050602
    Download Citation