Spectroscopy and Spectral Analysis, Volume. 42, Issue 2, 609(2022)

A Combination of Multiple Deep Learning Methods Applied to Small-Sample Space Objects Classification

Shi-yu DENG1、*, Cheng-zhi LIU1、1; 4; *;, Yong TAN3、3; *;, De-long LIU1、1;, Nan ZHANG1、1;, Zhe KANG1、1;, Zhen-wei LI1、1;, Cun-bo FAN1、1; 4;, Chun-xu JIANG3、3;, and Zhong LÜ3、3;
Author Affiliations
  • 11. Changchun Observatory of National Astronomical Observators, Chinese Academy of Sciences, Changchun 130117, China
  • 33. School of Science, Changchun University of Science and Technology, Changchun 130022, China
  • show less
    Figures & Tables(14)
    Schematic diagrams of optical telescope (a) and terminal box (b)
    Hyperspectral images of a space object at different wavelengths
    The brightness of space object varying with wavelength
    The basic architecture diagram of DBSCAN
    The basic flow chart of generative adversarial network
    The original and generated spectra of a space object
    The basic flow chart of one dimensional CNN
    The basic flow chart of experimental algorithm
    Comparisons of average operation time and accuracy among various methods
    Accuracies of four combination methods
    Average operation time of four combination methods
    • Table 1. Rough classification results based on DBSCAN

      View table
      View in Article

      Table 1. Rough classification results based on DBSCAN

      类型数量精度/%
      One5276.109 2
      Two975.880 9
      Three3270.901 4
      Four4275.910 1
      Five1166.666 7
      Six37-
    • Table 2. Rough classification results based on K-means

      View table
      View in Article

      Table 2. Rough classification results based on K-means

      类型数量精度/%
      One5775.503 7
      Two1075.900 9
      Three3770.001 5
      Four4574.999 9
      Five1466.798 1
      Six17-
    • Table 3. Classification accuracies (%) of six categories by using multiple methods combination

      View table
      View in Article

      Table 3. Classification accuracies (%) of six categories by using multiple methods combination

      类型方法
      DBGANCNNKMGANCNNDBOSCNNKMOSCNN
      One81.276 179.910 480.673 979.102 3
      Two80.890 180.125 680.110 179.912 2
      Three75.692 475.001 975.515 274.999 9
      Four80.902 980.511 180.190 780.111 3
      Five70.834 267.335 369.244 065.338 2
      Six85.000 184.600 984.859 383.981 5
    Tools

    Get Citation

    Copy Citation Text

    Shi-yu DENG, Cheng-zhi LIU, Yong TAN, De-long LIU, Nan ZHANG, Zhe KANG, Zhen-wei LI, Cun-bo FAN, Chun-xu JIANG, Zhong LÜ. A Combination of Multiple Deep Learning Methods Applied to Small-Sample Space Objects Classification[J]. Spectroscopy and Spectral Analysis, 2022, 42(2): 609

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: Jan. 8, 2021

    Accepted: Feb. 7, 2021

    Published Online: Apr. 2, 2022

    The Author Email: DENG Shi-yu (dengsy@cho.ac.cn)

    DOI:10.3964/j.issn.1000-0593(2022)02-0609-07

    Topics