• Infrared and Laser Engineering
  • Vol. 49, Issue 5, 20201010 (2020)
Zhang Panpan1,2,3,4,5,*, Luo Haibo1,2,4,5, Ju Moran1,2,3,4,5, Hui Bin1,2,4,5, and Chang Zheng1,2,4,5
Author Affiliations
  • 1[in Chinese]
  • 2[in Chinese]
  • 3[in Chinese]
  • 4[in Chinese]
  • 5[in Chinese]
  • show less
    DOI: 10.3788/irla20201010 Cite this Article
    Zhang Panpan, Luo Haibo, Ju Moran, Hui Bin, Chang Zheng. An improved Capsule and its application in target recognition of SAR images[J]. Infrared and Laser Engineering, 2020, 49(5): 20201010 Copy Citation Text show less
    Structure of Capsule unit
    Fig. 1. Structure of Capsule unit
    Layers of reconstruction of Capsule network
    Fig. 2. Layers of reconstruction of Capsule network
    Structure of original Capsule network and improved Capsule network
    Fig. 3. Structure of original Capsule network and improved Capsule network
    Optical images and their corresponding MSTAR SAR images for (a) and (b) BMP2, BTR70, T72, BTR60, and 2S1; (c) and (d) BRDM2, D7, T62, ZIL131, and ZSU23/4
    Fig. 4. Optical images and their corresponding MSTAR SAR images for (a) and (b) BMP2, BTR70, T72, BTR60, and 2S1; (c) and (d) BRDM2, D7, T62, ZIL131, and ZSU23/4
    Reconstruction result of improved Capsule. (a) Original image,(b) Target image and (c) Reconstruction image
    Fig. 5. Reconstruction result of improved Capsule. (a) Original image,(b) Target image and (c) Reconstruction image
    Reconstruction error curve and training loss curve of improved Capsule network. (a) Reconstruction error curve and (b) training loss curve
    Fig. 6. Reconstruction error curve and training loss curve of improved Capsule network. (a) Reconstruction error curve and (b) training loss curve
    Model size (parameters)Training_time/epochBFLOPs
    Capsule33.73 M2 min 8 s33.519
    Improved Capsule21.65 M1 min 3 s1.078
    Table 1. Performance comparison of improved Capsule network and Capsule network
    ClassBMP2sn-9563BTR70T72sn-132BTR602S1BRDM2D7T62ZIL131ZSU23/4
    Train samples( ) 117117116128150149150150150150
    Test samples( ) 195196196195274274274273274274
    Table 2. Raw SAR dataset for training and testing in experiment
    ClassBMP2sn-9563BTR70T72sn-132BTR602S1BRDM2D7T62ZIL131ZSU23/4
    BMP2sn-956396.920.512.570000000
    BTR700100.0000000000
    T72sn-13200100.000000000
    BTR6000098.4600.510001.03
    2S10002.9294.161.4600.730.730
    BRDM200.36500.73097.45001.090.365
    D700.73000099.27000
    T6200000.730098.9000.37
    ZIL13100000000100.000
    ZSU23/40000000.360099.64
    Table 3. Confusion matrix of 10-class target recognition results of Capsule network(recognition rate: 98.48%)
    ClassBMP2sn-9563BTR70T72sn-132BTR602S1BRDM2D7T62ZIL131ZSU23/4
    BMP2sn-956396.4103.590000000
    BTR700100.0000000000
    T72sn-13200100.000000000
    BTR6000098.97000001.03
    2S10002.55596.351.0950000
    BRDM20000.73098.540.36500.3650
    D70.36500000.36599.27000
    T62000000099.6300.37
    ZIL1310000000.36099.640
    ZSU23/40000000.360099.64
    Table 4. Confusion matrix of 10-class recognition results of improved Capsule network (recognition rate: 98.85%)
    MethodsSOC
    RatesTraining images
    SVM[7]90.10%3 670
    AdaBoost[7]92.70%3 670
    DCNN[9]92.30%3 671
    DCNN[8]94.56%2 747
    IGT[7]95.00%3 670
    CGM[10]97.18%3 670
    2-VDCNN[11]97.81%1 377
    CapsNet98.48%1 377
    Improved CapsNet98.85%1 377
    Table 5. Recognition performance of different methods
    Zhang Panpan, Luo Haibo, Ju Moran, Hui Bin, Chang Zheng. An improved Capsule and its application in target recognition of SAR images[J]. Infrared and Laser Engineering, 2020, 49(5): 20201010
    Download Citation