• Acta Photonica Sinica
  • Vol. 51, Issue 2, 0230003 (2022)
Yu FAN1, Huiqin WANG1、*, Ke WANG1, Zhan WANG2, and Gang ZHEN2
Author Affiliations
  • 1School of Information and Control Engineering,Xi'an University of Architecture and Technology,Xi'an 710055,China
  • 2Shaanxi Institute for the Preservation of Cultural Heritage,Xi'an 710075,China
  • show less
    DOI: 10.3788/gzxb20225102.0230003 Cite this Article
    Yu FAN, Huiqin WANG, Ke WANG, Zhan WANG, Gang ZHEN. Multi-output Least-squares SVR Spectral Reflectance Reconstruction Method Based on Adaptive Optimization in Multi-scene[J]. Acta Photonica Sinica, 2022, 51(2): 0230003 Copy Citation Text show less
    Structure diagram of MLS-SVR parameters adaptive optimization spectral reconstruction algorithm
    Fig. 1. Structure diagram of MLS-SVR parameters adaptive optimization spectral reconstruction algorithm
    The relationship between the number of training samples and the average spectral error
    Fig. 2. The relationship between the number of training samples and the average spectral error
    Comparative test sample RMSE
    Fig. 3. Comparative test sample RMSE
    Spectral reflectance curves of four reconstruction methods
    Fig. 4. Spectral reflectance curves of four reconstruction methods
    Color block markers for murals and painted cultural relics
    Fig. 5. Color block markers for murals and painted cultural relics
    Schematic diagram of working mode of optical fiber spectrometer and multispectral imaging system
    Fig. 6. Schematic diagram of working mode of optical fiber spectrometer and multispectral imaging system
    Schematic diagram of corresponding acquisition area between optical fiber spectrometer and multispectral imaging system
    Fig. 7. Schematic diagram of corresponding acquisition area between optical fiber spectrometer and multispectral imaging system
    Flow chart of spectral reconstruction
    Fig. 8. Flow chart of spectral reconstruction
    Relationship between the number of mixed training samples and the average spectral error
    Fig. 9. Relationship between the number of mixed training samples and the average spectral error
    Relationship between adaptive weight k1 and average spectral error
    Fig. 10. Relationship between adaptive weight k1 and average spectral error
    Comparison of reconstructed spectral curve and real spectral curve of reference color block
    Fig. 11. Comparison of reconstructed spectral curve and real spectral curve of reference color block
    The chromaticity distribution space of CIELAB reconstructed from reference color block by each reconstruction method
    Fig. 12. The chromaticity distribution space of CIELAB reconstructed from reference color block by each reconstruction method
    Evaluation standardValueReconstruction method
    Pseudo-inverseMSVRMLS-SVRSSA-MLS-SVR
    MRMSEMean0.01520.11390.01120.0084
    Min0.00470.00360.00390.0010
    Max0.04930.03800.03730.0370
    MGFC/%Mean97.9099.7299.7399.81
    Min86.0198.8398.8898.99
    Max99.9099.9799.9799.98
    ΔEMean5.0033.1753.0312.509
    Min1.4281.0440.9650.155
    Max8.3887.5106.5645.959
    Table 1. Accuracy comparison of four spectral reconstruction methods
    EvaluationArea codeReconstruction method
    Pseudo-inverseMSVRMLS-SVRSSA-MLS-SVR
    MRMSE10.03320.02760.02500.0121
    20.03430.01940.01630.0140
    30.03110.01900.01620.0085
    40.06260.03670.03280.0155
    50.06970.02200.01870.0087
    Mean0.04620.02490.02180.0018
    MGFC/%199.6599.7599.7999.95
    299.7799.9399.9499.96
    395.8499.6699.8199.89
    499.5699.8399.8699.95
    586.9199.6899.7699.95
    Mean96.3499.7799.8399.94
    ΔE14.8136.0405.3523.375
    25.6743.1012.7092.038
    39.0993.6692.8561.786
    49.3305.3484.8092.693
    516.1383.4892.9931.664
    Mean9.0114.3293.7442.311
    Table 2. Spectral reflectance reconstruction evaluation of 5 reference color blocks
    Area codeOriginalReconstruction method
    Pseudo-inverseMSVRMLS-SVRSSA-MLS-SVR
    1
    2
    3
    4
    5
    Table 3. Comparison of color restoration of reference color blocks reconstructed by different reconstruction methods
    Yu FAN, Huiqin WANG, Ke WANG, Zhan WANG, Gang ZHEN. Multi-output Least-squares SVR Spectral Reflectance Reconstruction Method Based on Adaptive Optimization in Multi-scene[J]. Acta Photonica Sinica, 2022, 51(2): 0230003
    Download Citation