• Journal of Infrared and Millimeter Waves
  • Vol. 39, Issue 4, 513 (2020)
Han-Lu ZHU1, Xu-Zhong ZHANG2, Xin CHEN3, Ting-Liang HU3, and Peng RAO3、*
Author Affiliations
  • 1Key Laboratory of Intelligent Infrared Perception, Chinese Academy of Sciences, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai200083, China
  • 2Huzhou Center for Applied Technology Research and Industrialization, Chinese Academy of Sciences, Huzhou1000, China
  • 3Key Laboratory of Intelligent Infrared Perception, Chinese Academy of Sciences, Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai200083, China
  • show less
    DOI: 10.11972/j.issn.1001-9014.2020.04.016 Cite this Article
    Han-Lu ZHU, Xu-Zhong ZHANG, Xin CHEN, Ting-Liang HU, Peng RAO. Dim small targets detection based on horizontal-vertical multi-scale grayscale difference weighted bilateral filtering[J]. Journal of Infrared and Millimeter Waves, 2020, 39(4): 513 Copy Citation Text show less
    [in Chinese]
    Fig. 1. [in Chinese]
    The response of different boundaries at different scales
    Fig. 2. The response of different boundaries at different scales
    calculate the HV-MSGD weighting operator
    Fig. 3. calculate the HV-MSGD weighting operator
    The detection process of the entire dim small target
    Fig. 4. The detection process of the entire dim small target
    Flow chart of trajectory detection
    Fig. 5. Flow chart of trajectory detection
    the results of five images that processed by our method, (a) is the input image, (b) is the 3D view of the input image, (c) is the image processed by HV-MSGD, (d) is a 3D image processed by HV-MSGD, (e) is a threshold segmentation image processed by (c), (f) is a 3D threshold segmentation image processed by (c)
    Fig. 6. the results of five images that processed by our method, (a) is the input image, (b) is the 3D view of the input image, (c) is the image processed by HV-MSGD, (d) is a 3D image processed by HV-MSGD, (e) is a threshold segmentation image processed by (c), (f) is a 3D threshold segmentation image processed by (c)
    Background suppression results of different algorithms in different scenarios (a) original image, (b) BF filtering result, (c) TDLMS filtering result in literature 15, (d) PM filtering result, (e) LCM filtering result of in literature 24, (f) NWIE, (g) filtering result of our method
    Fig. 7. Background suppression results of different algorithms in different scenarios (a) original image, (b) BF filtering result, (c) TDLMS filtering result in literature 15, (d) PM filtering result, (e) LCM filtering result of in literature 24, (f) NWIE, (g) filtering result of our method
    ROC of the five sequences
    Fig. 8. ROC of the five sequences
    Track of five sequences
    Fig. 9. Track of five sequences
    Histograms of detected bias pixels obtained by using our method (a) histograms of horizontal detected bias pixels of five sequences, (b) histograms of vertical detected bias pixels of five sequences
    Fig. 10. Histograms of detected bias pixels obtained by using our method (a) histograms of horizontal detected bias pixels of five sequences, (b) histograms of vertical detected bias pixels of five sequences
    ImageInputBFTDLMSPM
    SNRinσinBSFGSNRBSFGSNRBSFGSNR
    Image 12.7830.052.156.568.198.51.733.26
    Image 21.7742.874.0813.1218.727.981.333.12
    Image 31.7959.951.824.336.374.571.774.33
    Image 41.1336.731.388.1911.774.920.691.5
    Image 51.1629.267.120.8931.4614.242.567.12
    ImageLCMNWIEOur method
    BSFGSNRBSFGSNRBSFGSNR
    Image 12.637.74.551.5913.4840.99
    Image 22.638.894.4318.9421.3371.37
    Image 31.612.982.7216.9811.7327.53
    Image 41.091.928.467.8220.6312.65
    Image 54.414.5921.5221.32121.92131
    Table 1. BSF and GSNR of different five images processed by different methods
    IMAGEFrame numberInputBFTDLMSPM
    σin¯SNRin¯BSF¯GSNR¯BSF¯GSNR¯BSF¯GSNR¯
    Seq 17525.722.233.134.361.013.583.12.31
    Seq 2908.360.561.225.722.425.121.284.56
    Seq 36561.219.310.682.558.246.167.37
    Seq 49032.782.092.649.495.717.611.194.01
    Seq 59011.531.171.076.771.24.861.953.89
    Average-17.221.443.127.322.645.822.514.33
    IMAGEFrame numberLCMNWIEOur method
    BSF¯GSNR¯BSF¯GSNR¯BSF¯GSNR¯
    Seq 1753.11.271.986.045.3833.97
    Seq 2901.012.022.409.425.4615.64
    Seq 3654.763.131.3413.7314.7155.18
    Seq 4901.177.943.1710.2523.2681.8
    Seq 5902.941.611.7413.17.4747.51
    Average-2.453.272.1310.5111.2636.47
    Table 2. the average of BSF and GSNR for five sequences
    Pd

    Seq 1

    SNRin¯=2.23

    Seq2

    SNRin¯=0.56

    Seq 3

    SNRin¯=1.21

    Seq 4

    SNRin¯=2.09

    Seq 5

    SNRin¯=1.17

    Average

    SNRin¯=1.44

    BF0.92000.94700.87040.92530.95250.9230
    TDLMS0.72020.74270.43420.47650.93340.6614
    LCM0.63740.84390.28710.90560.67380.6696
    PM0.55510.76450.30610.85770.63380.6234
    NWIE0.84770.91670.55780.86140.95070.8269
    Our method0.93200.95770.93490.98470.97610.9571
    Table 3. AUC of five sequences in different algorithms
    Han-Lu ZHU, Xu-Zhong ZHANG, Xin CHEN, Ting-Liang HU, Peng RAO. Dim small targets detection based on horizontal-vertical multi-scale grayscale difference weighted bilateral filtering[J]. Journal of Infrared and Millimeter Waves, 2020, 39(4): 513
    Download Citation