• Acta Photonica Sinica
  • Vol. 51, Issue 12, 1206002 (2022)
Yang CAO, Zupeng ZHANG*, and Xiaofeng PENG
Author Affiliations
  • School of Electrical and Electronic Engineering,Chongqing University of Technology,Chongqing 400054,China
  • show less
    DOI: 10.3788/gzxb20225112.1206002 Cite this Article
    Yang CAO, Zupeng ZHANG, Xiaofeng PENG. Wavefront Distortion Restoration Method Based on Residual Attention Network[J]. Acta Photonica Sinica, 2022, 51(12): 1206002 Copy Citation Text show less
    Schematic diagram of AO system without wavefront detection
    Fig. 1. Schematic diagram of AO system without wavefront detection
    Overall network structure
    Fig. 2. Overall network structure
    Structure diagram of mixed attention
    Fig. 3. Structure diagram of mixed attention
    Turbulent phase distribution and light intensity distribution
    Fig. 4. Turbulent phase distribution and light intensity distribution
    Loss function and accuracy of the model
    Fig. 5. Loss function and accuracy of the model
    Comparison of the predicted turbulence phase with the actual phase
    Fig. 6. Comparison of the predicted turbulence phase with the actual phase
    Comparison of the predicted Zernike coefficient with the actual coefficient
    Fig. 7. Comparison of the predicted Zernike coefficient with the actual coefficient
    Light intensity map at different SNR
    Fig. 8. Light intensity map at different SNR
    Residual phase at different SNR
    Fig. 9. Residual phase at different SNR
    Comparison of prediction results between the model with partial attention mechanism removed and the complete model
    Fig. 10. Comparison of prediction results between the model with partial attention mechanism removed and the complete model
    Comparison of partial loss function with complete model
    Fig. 11. Comparison of partial loss function with complete model
    ParameterValue
    Laser wavelength λ1 550 nm
    Width of phase screen D0.3 m
    Beam waist w03 cm
    Topological charge l3
    Radial index p0
    Transmission distance z1 km
    Number of phase screens10
    Table 1. Parameter of simulation
    D/r0=2D/r0=5D/r0=10D/r0=15D/r0=20
    OursRef.[12OursRef.[12OursRef.[12OursRef.[12OursRef.[12
    PV/rad0.0450.0820.1950.2140.2260.4150.1960.4390.2750.381
    RMS/rad0.0110.0240.0320.0810.0460.1690.0530.2250.0710.315
    Table 2. PV and RMS values of the residual phase at different turbulence intensities
    SNRD/r0=2D/r0=5D/r0=10D/r0=15D/r0=20
    PV/radRMS/radPV/radRMS/radPV/radRMS/radPV/radRMS/radPV/radRMS/rad
    5 dB0.1310.0230.1780.0510.4810.1330.7310.1520.6240.129
    15 dB0.0510.0110.1350.0300.2440.1060.3910.1250.3420.095
    25 dB0.0310.0060.1050.0150.1380.0690.1380.1180.1210.071
    Table 3. PV and RMS values of the residual phase at different SNR
    ModelAccuracyTime/ms
    ResNet500.8928.35
    ResNet50+SA0.9349.07
    ResNet50+CA0.9419.44
    ResNet50+CA+SA0.9719.51
    Table 4. Comparison of accuracy and calculation time of different models
    Complete modelRemove CARemove SA
    PV/rad0.145 10.258 90.373
    RMS/rad0.052 70.077 10.088 1
    Table 5. Comparison of evaluation indexes of removing partial attention mechanism
    Complete modelRemove PVRemove RMS
    PV/rad0.164 90.354 60.274 7
    RMS/rad0.039 50.070 80.080 9
    Table 6. Comparison of evaluation indexes of partial loss function
    Yang CAO, Zupeng ZHANG, Xiaofeng PENG. Wavefront Distortion Restoration Method Based on Residual Attention Network[J]. Acta Photonica Sinica, 2022, 51(12): 1206002
    Download Citation