• Laser & Optoelectronics Progress
  • Vol. 57, Issue 22, 223001 (2020)
Jun Hu, Zhen Xu, Maopeng Li, and Yande Liu*
Author Affiliations
  • School of Mechatronics & Vehicle Engineering, East China Jiaotong University, Nanchang, Jiangxi 330013, China
  • show less
    DOI: 10.3788/LOP57.223001 Cite this Article Set citation alerts
    Jun Hu, Zhen Xu, Maopeng Li, Yande Liu. Determination of Melamine Content in Milk Powder Based on Neural Network Algorithm and Terahertz Spectrum Detection[J]. Laser & Optoelectronics Progress, 2020, 57(22): 223001 Copy Citation Text show less
    Topology structure of BPNN
    Fig. 1. Topology structure of BPNN
    Architecture of GRNN
    Fig. 2. Architecture of GRNN
    Terahertz absorbance of melamine adulterated milk powder samples with different mass concentrations
    Fig. 3. Terahertz absorbance of melamine adulterated milk powder samples with different mass concentrations
    Refractive index spectra of pure milk powder and pure melamine, and refractive index spectra of mixed samples. (a) Refractive index spectra of pure milk powder and pure melamine; (b) refractive index spectra of mixed samples
    Fig. 4. Refractive index spectra of pure milk powder and pure melamine, and refractive index spectra of mixed samples. (a) Refractive index spectra of pure milk powder and pure melamine; (b) refractive index spectra of mixed samples
    Optimized results of number of input vectors and number of hidden layers for BPNN
    Fig. 5. Optimized results of number of input vectors and number of hidden layers for BPNN
    Predicted accuracy comparison of GRNN model with different numbers of input vectors
    Fig. 6. Predicted accuracy comparison of GRNN model with different numbers of input vectors
    Predicted accuracy comparison of GRNN model with different spread values
    Fig. 7. Predicted accuracy comparison of GRNN model with different spread values
    Predicted results of BPNN and GRNN models for mass concentration of melamine
    Fig. 8. Predicted results of BPNN and GRNN models for mass concentration of melamine
    DatasetNumber ofsamplesMass concentration /%
    MinimumMaximumMeanStandard deviation
    Total1760.04319.9889.3176.261
    Modeling1320.04319.9889.3976.311
    Prediction440.04319.9889.0696.169
    Table 1. True value distribution of modeling and prediction datasets of melamine doped milk powder with different mass concentrations
    ModelPreprocessing methodPCCalibrationPrediction
    rcRMSECrpRMSEP
    PLSNone70.98810.00980.97520.0127
    Savitzky-Golay70.98890.00950.97620.0124
    MSC20.99010.00890.98140.0112
    Baseline correction40.97970.01280.95930.0165
    Normalization30.98970.00910.98130.0112
    MSC+Normalization30.99120.00850.98420.0103
    Table 2. Results of correction processing for THz absorbance of melamine doped milk powder
    Jun Hu, Zhen Xu, Maopeng Li, Yande Liu. Determination of Melamine Content in Milk Powder Based on Neural Network Algorithm and Terahertz Spectrum Detection[J]. Laser & Optoelectronics Progress, 2020, 57(22): 223001
    Download Citation