• Acta Photonica Sinica
  • Vol. 49, Issue 3, 0330001 (2020)
Xin WANG1、2, Zhe-ming KANG1, Long LIU1, and Xian-guang FAN1、2、*
Author Affiliations
  • 1Department of Instrumental and Electrical Engineering, Xiamen University, Xiamen, Fujian 361005, China
  • 2Fujian Key Laboratory of Universities and Colleges for Transducer Technology, Xiamen Key Laboratory of Optoelectronic Transducer Technology, Xiamen, Fujian 361005, China
  • show less
    DOI: 10.3788/gzxb20204903.0330001 Cite this Article
    Xin WANG, Zhe-ming KANG, Long LIU, Xian-guang FAN. Multi-channel Raman Spectral Reconstruction Based on Gaussian Kernel Principal Component Analysis[J]. Acta Photonica Sinica, 2020, 49(3): 0330001 Copy Citation Text show less
    Process of Raman spectral reconstruction algorithm based on KPCA
    Fig. 1. Process of Raman spectral reconstruction algorithm based on KPCA
    The calibration Raman spectra of PMMA measured by different integration times and sample preparation methods
    Fig. 2. The calibration Raman spectra of PMMA measured by different integration times and sample preparation methods
    The Raman spectrum of PMMA with random fluorescence background and noise and the Raman spectrum of PMMA without fluorescence background and noise
    Fig. 3. The Raman spectrum of PMMA with random fluorescence background and noise and the Raman spectrum of PMMA without fluorescence background and noise
    Full reconstructed Raman spectra based on PINV, Wiener estimation, PCA and KPCA, respectively
    Fig. 4. Full reconstructed Raman spectra based on PINV, Wiener estimation, PCA and KPCA, respectively
    Multi-channel narrow-band image
    Fig. 5. Multi-channel narrow-band image
    The Raman images reconstructed by PINV, Wiener estimation and KPCA, respectively
    Fig. 6. The Raman images reconstructed by PINV, Wiener estimation and KPCA, respectively
    Full Raman spectra based on pixel 1 and 2 reconstructed by PINV, Wiener estimation and KPCA, respectively
    Fig. 7. Full Raman spectra based on pixel 1 and 2 reconstructed by PINV, Wiener estimation and KPCA, respectively
    Raman spectrar0r1r2r3r4
    Integration times10 s3 s10 s10 s10 s
    Preparation methodsOriginal packageOriginal packageValve bagQuartz tubeGlass tube
    Table 1. The calibration Raman spectra of PMMA measured by different integration times and sample preparation methods
    Methodr0r1r2r3r4Mean
    PINV0.015 90.015 80.017 80.008 50.018 60.015 3
    Wiener estimation0.014 20.013 90.010 30.012 90.019 00.014 1
    PCA0.009 90.010 90.011 00.010 80.009 10.010 4
    KPCA0.005 00.004 20.006 90.006 70.015 70.007 7
    Table 2. RMSEs offull reconstructed Raman spectra based on PINV, wiener estimation, PCA and KPCA
    MethodSNRPINVWiener estimationPCAKPCA
    Test 113.072 60.015 30.014 10.010 40.007 7
    Test 210.614 00.021 80.016 40.010 10.007 8
    Test 39.637 30.029 70.018 60.011 00.007 6
    Test 48.236 50.037 10.021 30.012 00.008 0
    Test 57.770 50.042 70.024 70.014 70.009 7
    Test 67.457 40.048 60.028 80.018 70.013 5
    Test 76.702 90.055 50.033 40.026 40.016 3
    Test 85.921 40.064 20.039 30.032 50.021 3
    Test 95.506 30.075 10.046 90.041 50.029 7
    Test 105.121 00.088 10.058 60.055 50.044 3
    Table 3. RMSEs of full reconstructed Raman spectra based on PINV, wiener estimation, PCA and KPCA under conditions of different noise
    MethodPINVWiener estimationKPCA
    Pixel 10.032 90.018 20.008 4
    Pixel 20.042 80.022 50.011 1
    Table 4. RMSEs of full Raman spectra based on pixel 1 and pixel 2 reconstructed by PINV, Wiener estimation and KPCA, respectively
    Xin WANG, Zhe-ming KANG, Long LIU, Xian-guang FAN. Multi-channel Raman Spectral Reconstruction Based on Gaussian Kernel Principal Component Analysis[J]. Acta Photonica Sinica, 2020, 49(3): 0330001
    Download Citation