• Infrared and Laser Engineering
  • Vol. 49, Issue 10, 20200221 (2020)
Xu Zhang, Mingxin Yu, Lianqing Zhu, Yanlin He, and Guangkai Sun
Author Affiliations
  • Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing Information Science and Technology University, Beijing 100016, China
  • show less
    DOI: 10.3788/IRLA20200221 Cite this Article
    Xu Zhang, Mingxin Yu, Lianqing Zhu, Yanlin He, Guangkai Sun. Raman mineral recognition method based on all-optical diffraction deep neural network[J]. Infrared and Laser Engineering, 2020, 49(10): 20200221 Copy Citation Text show less
    Part of the spectrum data in dataset
    Fig. 1. Part of the spectrum data in dataset
    Network structure diagram
    Fig. 2. Network structure diagram
    Data preprocessing
    Fig. 3. Data preprocessing
    Detection area in the result image and identification area of different labels
    Fig. 4. Detection area in the result image and identification area of different labels
    Structure diagram of optical diffractive neural network classification system
    Fig. 5. Structure diagram of optical diffractive neural network classification system
    Structure diagram of grating group
    Fig. 6. Structure diagram of grating group
    Grating height discretization
    Fig. 7. Grating height discretization
    Two etching to achieve four steps
    Fig. 8. Two etching to achieve four steps
    Partial diagram of diffraction grating model
    Fig. 9. Partial diagram of diffraction grating model
    Accuracy and error curve of train set
    Fig. 10. Accuracy and error curve of train set
    Accuracy curve of train set and test set
    Fig. 11. Accuracy curve of train set and test set
    Confusion matrix of test results
    Fig. 12. Confusion matrix of test results
    Height distribution of each layer of grating and the output image of each layer of grating after training. (a1)-(a6) Gratings; (b1)-(b6) Output of gratings; (c1)-(c3) Recognition
    Fig. 13. Height distribution of each layer of grating and the output image of each layer of grating after training. (a1)-(a6) Gratings; (b1)-(b6) Output of gratings; (c1)-(c3) Recognition
    Height precisionAccuracyLoss of accuracy
    Original data94.220%0
    6 bit93.642%0.613%
    5 bit93.064%1.227%
    4 bit90.751%3.682%
    3 bit86.705%7.976%
    2 bit63.584%32.513%
    1 bit15.222%83.844%
    Table 1. [in Chinese]
    Xu Zhang, Mingxin Yu, Lianqing Zhu, Yanlin He, Guangkai Sun. Raman mineral recognition method based on all-optical diffraction deep neural network[J]. Infrared and Laser Engineering, 2020, 49(10): 20200221
    Download Citation