• Laser & Optoelectronics Progress
  • Vol. 59, Issue 8, 0830001 (2022)
Mengran Zhou, Rongying Dai*, Chen Yang, Feng Hu, Kai Bian, Wenhao Lai, and Xixi Kong
Author Affiliations
  • College of Electrical and Information Engineering, Anhui University of Science and Technology, Huainan , Anhui 232001, China
  • show less
    DOI: 10.3788/LOP202259.0830001 Cite this Article Set citation alerts
    Mengran Zhou, Rongying Dai, Chen Yang, Feng Hu, Kai Bian, Wenhao Lai, Xixi Kong. Fast Nondestructive Detection of Edible Oil Based on Fluorescence Spectrum and Stack Autoencoder[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0830001 Copy Citation Text show less
    SAE network architecture
    Fig. 1. SAE network architecture
    Diagram of experimental equipment for LIF
    Fig. 2. Diagram of experimental equipment for LIF
    Original fluorescence spectra of different edible oil samples. (a) All samples; (b) sample of rapeseed oil; (c) sample of corn oil; (d) sample of peanut oil; (e) mixed oil sample A; (f) mixed oil sample B
    Fig. 3. Original fluorescence spectra of different edible oil samples. (a) All samples; (b) sample of rapeseed oil; (c) sample of corn oil; (d) sample of peanut oil; (e) mixed oil sample A; (f) mixed oil sample B
    Relationship between number of hidden layers and Softmax classification accuracy
    Fig. 4. Relationship between number of hidden layers and Softmax classification accuracy
    Effect of hidden layer neurons on ELM performance
    Fig. 5. Effect of hidden layer neurons on ELM performance
    Classification results of spectrum test set processed by SAE
    Fig. 6. Classification results of spectrum test set processed by SAE
    Effect of hidden layer neurons on ELM performance
    Fig. 7. Effect of hidden layer neurons on ELM performance
    Independent validation of classification results of set
    Fig. 8. Independent validation of classification results of set
    LabelTypeVolume ratio
    1Rapeseed oil
    2Corn oil

    3

    4

    5

    Peanut oil

    Peanut oil mixed with rapeseed oil

    Peanut oil mixed with corn oil

    1∶1

    1∶1

    Table 1. Experimental sample materials

    Experimental

    group

    Number of first hidden layerNumber of second hidden layerNumber of third hidden layerNumber of fourth hidden layer
    1300
    230050
    33005010
    430050105
    Table 2. Number of hidden layers in each experimental group
    ModelNumber of hidden layersMean accuracy /%Variance
    SAE 1175.43.84
    SAE 2298.40.64
    SAE 3385.43.84
    SAE 4480.65.04
    Table 3. Average classification accuracy and variance under different number of hidden layers
    Modeling methodAccuracy of training set /%Accuracy of test set /%
    SAE+BP100(400/400)100(100/100)
    SAE+ELM100(400/400)100(100/100)
    SAE+SVM100(400/400)100(100/100)

    SAE+decision tree

    PCA+BP

    PCA+ELM

    PCA+SVM

    PCA+decision tree

    100(400/400)

    94(376/400)

    97(388/400)

    97(386/400)

    100(400/400)

    100(100/100)

    97(97/100)

    99(99/100)

    97(97/100)

    97(97/100)

    Table 4. Classification results for different recognition models
    Modeling methodSAE+BPSAE+ELMSAE+SVMSAE+decision tree
    Testing network time197.50.21.616.4
    Table 5. Test network time for different recognition models
    Mengran Zhou, Rongying Dai, Chen Yang, Feng Hu, Kai Bian, Wenhao Lai, Xixi Kong. Fast Nondestructive Detection of Edible Oil Based on Fluorescence Spectrum and Stack Autoencoder[J]. Laser & Optoelectronics Progress, 2022, 59(8): 0830001
    Download Citation