• Spectroscopy and Spectral Analysis
  • Vol. 42, Issue 1, 292 (2022)
Qing-liang JIAO1、*, Ming LIU1、1; *;, Kun YU2、2;, Zi-long LIU2、2; 3;, Ling-qin KONG1、1;, Mei HUI1、1;, Li-quan DONG1、1;, and Yue-jin ZHAO1、1;
Author Affiliations
  • 11. Beijing Key Laboratory for Precision Optoelectronic Measurement Instrument and Technology, School of Optics and Photonics, Beijing Institute of Technology, Beijing 100081, China
  • 22. Henan Key Laboratory of Infrared Materials and Spectrum Measures and Applications, School of Physics, Henan Normal University, Xinxiang 453007, China
  • show less
    DOI: 10.3964/j.issn.1000-0593(2022)01-0292-06 Cite this Article
    Qing-liang JIAO, Ming LIU, Kun YU, Zi-long LIU, Ling-qin KONG, Mei HUI, Li-quan DONG, Yue-jin ZHAO. Spectral Pre-Processing Based on Convolutional Neural Network[J]. Spectroscopy and Spectral Analysis, 2022, 42(1): 292 Copy Citation Text show less
    The structure of proposed CNN
    Fig. 1. The structure of proposed CNN
    The result of spectral denoising(a): RMSE; (b): GFC
    Fig. 2. The result of spectral denoising
    (a): RMSE; (b): GFC
    The result of baseline correction(a): RMSE; (b): GFC
    Fig. 3. The result of baseline correction
    (a): RMSE; (b): GFC
    Effect of denoising and baseline correction(a): Noise and Ours; (b): DWT+[6], DWT+[8] and Ours; (c): SG+[6], SG+[8] and Ouss
    Fig. 4. Effect of denoising and baseline correction
    (a): Noise and Ours; (b): DWT+[6], DWT+[8] and Ours; (c): SG+[6], SG+[8] and Ouss
    SNR of noiseDWT+[6]DWT+[8]SG+[6]SG+[8]Ours
    00.981\6.3790.973\5.9630.577\1.1970.571\1.1830.146\1.011
    100.313\1.9610.308\1.9640.203\0.9760.219\0.9710.059\0.980
    200.099\1.0610.096\1.0590.068\0.9520.067\0.9500.020\0.988
    300.031\1.0120.030\1.0100.011\1.0010.011\1.0010.006\0.999
    400.010\1.0000.010\1.0000.007\0.9980.007\0.9990.002\0.999
    Table 1. The result of denoising and baseline correction(RMSE\GFC)
    标准TD[10]Ours
    位置强度位置强度位置强度位置强度
    130.586 730.586 730.586 730.586 7
    270.843 970.843 970.843 970.843 9
    3131.055 413.261.030 713.141.053 1131.055 4
    4150.875 5--14.690.919 615.020.870 5
    5190.762 3190.762 3190.762 3190.762 3
    Table 2. The result of spectrum peak detection
    DenoisingDWTSG
    Baseline[6][8][6][8]Ours
    PeakTD[10]TD[10]TD[10]TD[10]
    Moisture0.1580.1480.1560.1470.1530.1320.1530.1360.119
    Oil0.7830.6300.7810.6330.7590.6160.7530.6110.414
    Table 3. RMSE of quantitative analysis
    Qing-liang JIAO, Ming LIU, Kun YU, Zi-long LIU, Ling-qin KONG, Mei HUI, Li-quan DONG, Yue-jin ZHAO. Spectral Pre-Processing Based on Convolutional Neural Network[J]. Spectroscopy and Spectral Analysis, 2022, 42(1): 292
    Download Citation