• Acta Optica Sinica
  • Vol. 37, Issue 12, 1201002 (2017)
Zhi Cheng1、2, Feng He1、*, Silong Zhang1, Xu Jing1, and Zaihong Hou1
Author Affiliations
  • 1 Key Laboratory of Atmospheric Composition and Optical Radiation, Chinese Academy of Sciences, Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Hefei, Anhui 230031, China
  • 2 University of Science and Technology of China, Hefei, Anhui 230026, China
  • show less
    DOI: 10.3788/AOS201737.1201002 Cite this Article Set citation alerts
    Zhi Cheng, Feng He, Silong Zhang, Xu Jing, Zaihong Hou. Combination Method of Wavelet and Empirical Mode Decomposition with Trend Modulation used for Atmospheric Coherent Length Profile Denoising[J]. Acta Optica Sinica, 2017, 37(12): 1201002 Copy Citation Text show less
    Schematic of measurement principle of differential light column image motion lidar
    Fig. 1. Schematic of measurement principle of differential light column image motion lidar
    Average SNR of r0 profile denoising with three wavelet bases versus number of decomposition layers under 5% Gauss noise level
    Fig. 2. Average SNR of r0 profile denoising with three wavelet bases versus number of decomposition layers under 5% Gauss noise level
    Decomposition results of noisy r0 profile with SNR of 12 dB
    Fig. 3. Decomposition results of noisy r0 profile with SNR of 12 dB
    DFA exponent of each layer of IMF
    Fig. 4. DFA exponent of each layer of IMF
    Denoised atmospheric coherent length profiles obtained by different methods under different Gauss noise levels. (a) 5% Gauss noise; (b) 10% Gauss noise; (c) 15% Gauss noise; (d) 20% Gauss noise
    Fig. 5. Denoised atmospheric coherent length profiles obtained by different methods under different Gauss noise levels. (a) 5% Gauss noise; (b) 10% Gauss noise; (c) 15% Gauss noise; (d) 20% Gauss noise
    Retrieved Cn2 profiles before and after denoising under different Gauss noise levels with different methods. (a) 5% Gauss noise; (b) 10% Gauss noise; (c) 15% Gauss noise; (d) 20% Gauss noise
    Fig. 6. Retrieved Cn2 profiles before and after denoising under different Gauss noise levels with different methods. (a) 5% Gauss noise; (b) 10% Gauss noise; (c) 15% Gauss noise; (d) 20% Gauss noise
    Undenoised and denoised atmospheric coherent length profiles of DCIM lidar at different time. (a) 19:59:29 on November 9, 2015; (b) 14:40:26 on November 11, 2015; (c) 20:34:50 on November 11, 2015
    Fig. 7. Undenoised and denoised atmospheric coherent length profiles of DCIM lidar at different time. (a) 19:59:29 on November 9, 2015; (b) 14:40:26 on November 11, 2015; (c) 20:34:50 on November 11, 2015
    Comparison between undenoised and denoised atmospheric coherent length profiles of DCIM lidar and balloon profile at different time. (a) 19:59:29 on November 9, 2015; (b) 14:40:26 on November 11, 2015; (c) 20:34:50 on November 12, 2015
    Fig. 8. Comparison between undenoised and denoised atmospheric coherent length profiles of DCIM lidar and balloon profile at different time. (a) 19:59:29 on November 9, 2015; (b) 14:40:26 on November 11, 2015; (c) 20:34:50 on November 12, 2015
    Method5% Gauss noise10% Gauss noise15% Gauss noise20% Gauss noise
    SNR /dBRMSE /dBSNR /dBRMSE /dBSNR /dBRMSE /dBSNR /dBRMSE /dB
    Wavelet35.04030.112230.38970.181227.36770.277125.43550.3653
    EMD30.50110.188125.32900.353621.33460.540619.59720.6803
    EEMD32.36560.152328.19770.224225.28320.358223.38920.4563
    Wavelet-EMD36.21720.109132.14520.196829.30740.222027.36420.2905
    Wavelet-trend-EMD37.55070.097233.27410.156331.13080.187428.91570.2492
    Table 1. Statistical results of SNR and root-mean-square error (RMSE) of denoised atmospheric coherent length profiles obtained by different methods under different Gauss noise levels
    Method5% Gauss noise10% Gauss noise15% Gauss noise20% Gauss noise
    SNR /dBRMSE /dBSNR /dBRMSE /dBSNR /dBRMSE /dBSNR /dBRMSE /dB
    Without denoising34.66620.326033.95780.372433.02010.410331.65660.4979
    Wavelet36.61170.258535.83950.302534.99850.317933.39220.3921
    EMD35.32180.308934.27280.358233.53350.379231.92870.4671
    EEMD5.79060.283434.38510.353334.14090.355632.14640.4489
    Wavelet-EMD37.09840.237336.41470.239535.34770.307534.46550.3501
    Wavelet-trend-EMD37.76160.204436.90760.214135.98320.289334.87230.3286
    Table 2. Statistical results of SNR and RMSE of Cn2 profile retrieved from undenoised and denoised r0 profile under different gauss noise levels
    MethodNovember 9, 2015November 11,2015November 12,2015
    SNR /dBRMSE /dBSNR /dBRMSE /dBSNR /dBRMSE /dB
    Without denoising29.77230.560330.63130.453129.57400.5570
    Wavelet30.11820.547731.28130.420534.99850.3179
    EMD29.92490.550530.99400.434630.19180.5187
    EEMD29.96320.543230.86630.441031.19190.4623
    Wavelet-EMD30.22450.538031.49120.410431.75270.4334
    Wavelet-trend-EMD30.89700.492231.87620.391832.07450.4176
    Table 3. Undenoised and denoised SNR and RMSE of retrieved Cn2 profiles measured by DCIM lidar
    Zhi Cheng, Feng He, Silong Zhang, Xu Jing, Zaihong Hou. Combination Method of Wavelet and Empirical Mode Decomposition with Trend Modulation used for Atmospheric Coherent Length Profile Denoising[J]. Acta Optica Sinica, 2017, 37(12): 1201002
    Download Citation