• Laser & Optoelectronics Progress
  • Vol. 55, Issue 11, 111504 (2018)
Hao Zhang** and Changhong Chen*
Author Affiliations
  • College of Communication and Information Technology, Nanjing University of Posts and Telecommunications, Nanjing, Jiangsu 210003, China
  • show less
    DOI: 10.3788/LOP55.111504 Cite this Article Set citation alerts
    Hao Zhang, Changhong Chen. Aurora Sequence Classification Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2018, 55(11): 111504 Copy Citation Text show less
    Framework of our method
    Fig. 1. Framework of our method
    Four categories of sample images at 557.7 nm
    Fig. 2. Four categories of sample images at 557.7 nm
    CNN with attribute constraints
    Fig. 3. CNN with attribute constraints
    Feature maps of facula attributes learned from our network. (a) Arc image; (b) hot-spot image; (c) arc feature map; (d) hot-spot feature map
    Fig. 4. Feature maps of facula attributes learned from our network. (a) Arc image; (b) hot-spot image; (c) arc feature map; (d) hot-spot feature map
    Classification accuracy of aurora images with different ω
    Fig. 5. Classification accuracy of aurora images with different ω
    Comparison of classification accuracy on aurora sequences
    Fig. 6. Comparison of classification accuracy on aurora sequences
    Distribution of four kinds of aurora
    Fig. 7. Distribution of four kinds of aurora
    MethodAllArcDraperyRadialHot-spot
    Original CNN0.950.980.890.970.90
    CNN-LSTM on parallel connectionOur method (ω=4)0.960.950.980.980.910.890.910.900.980.95
    Table 1. Comparison of classification accuracy on aurora images
    Hao Zhang, Changhong Chen. Aurora Sequence Classification Based on Deep Learning[J]. Laser & Optoelectronics Progress, 2018, 55(11): 111504
    Download Citation