• Laser & Optoelectronics Progress
  • Vol. 56, Issue 10, 101007 (2019)
Jingjing Sun1,2,** and Fei Zhao1,*
Author Affiliations
  • 1 Key Laboratory of Computational Optical Imaging Technology, Academy of Opto-Electronics, Chinese Academy of Sciences, Beijing 100094, China
  • 2 Univercity of Chinese Academy of Sciences, Beijing 100049, China
  • show less
    DOI: 10.3788/LOP56.101007 Cite this Article Set citation alerts
    Jingjing Sun, Fei Zhao. Application of Non-Negative Matrix Factorization in Space Object Recognition[J]. Laser & Optoelectronics Progress, 2019, 56(10): 101007 Copy Citation Text show less
    Principle diagram of NMF
    Fig. 1. Principle diagram of NMF
    Scanning mode
    Fig. 2. Scanning mode
    Geometric model of laboratory
    Fig. 3. Geometric model of laboratory
    Images of ten kinds of space objects
    Fig. 4. Images of ten kinds of space objects
    Flow chart of experiment. (a) Feature extraction; (b) space object image recognition
    Fig. 5. Flow chart of experiment. (a) Feature extraction; (b) space object image recognition
    Recognition rate of each NMF algorithm under number of samples is 100
    Fig. 6. Recognition rate of each NMF algorithm under number of samples is 100
    Training time of each NMF algorithm under number of samples is 100
    Fig. 7. Training time of each NMF algorithm under number of samples is 100
    Recognition rate of each NMF algorithm under different numbers of samples
    Fig. 8. Recognition rate of each NMF algorithm under different numbers of samples
    Training time of each NMF algorithm under different numbers of samples
    Fig. 9. Training time of each NMF algorithm under different numbers of samples
    AlgorithmEDNMFKLNMFSNMF2DED2DKL2DS(2D)2ED(2D)2KL(2D)2S
    Testing time /s0.9941.0141.0360.6620.6760.6650.7130.7340.722
    Table 1. Testing time of each NMF algorithm under number of samples is 100
    AlgorithmEDNMFKLNMFSNMF2DED2DKL2DS(2D)2ED(2D)2KL(2D)2S
    Testing time/s1.0311.0361.0140.5350.5480.5320.7220.7280.709
    Table 2. Testing time of each algorithm under optimal k (number of samples is 100)
    MethodNMFMethod inRef. [2]Method inRef. [3]Method inRef. [4]
    Recognitionrate0.90.720.840.67
    Trainingtime /s38.91135.475748.65500.43
    Table 3. Comparison between NMF and existing methods
    Jingjing Sun, Fei Zhao. Application of Non-Negative Matrix Factorization in Space Object Recognition[J]. Laser & Optoelectronics Progress, 2019, 56(10): 101007
    Download Citation