• Laser & Optoelectronics Progress
  • Vol. 58, Issue 8, 0810008 (2021)
Baodai Shi*, Qin Zhang, Yao Li, and Yuhuan Li
Author Affiliations
  • College of Graduate, Air Force Engineering University, Xi'an, Shaanxi 710051, China
  • show less
    DOI: 10.3788/LOP202158.0810008 Cite this Article Set citation alerts
    Baodai Shi, Qin Zhang, Yao Li, Yuhuan Li. SAR Image Target Recognition Based on Improved Residual Attention Network[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810008 Copy Citation Text show less
    10 types of target optical images and corresponding SAR images
    Fig. 1. 10 types of target optical images and corresponding SAR images
    ResNet101 precision curve
    Fig. 2. ResNet101 precision curve
    SENet structure
    Fig. 3. SENet structure
    Residual shrinkage block
    Fig. 4. Residual shrinkage block
    Structure diagram of model S
    Fig. 5. Structure diagram of model S
    Concrete structure of the first stage
    Fig. 6. Concrete structure of the first stage
    Occlusion effect pictures
    Fig. 7. Occlusion effect pictures
    Pictures of salt and pepper noise
    Fig. 8. Pictures of salt and pepper noise
    CategoryNumber of samples in training setNumber of samples in test set
    2S1299274
    BMP2232196
    BRDM-2298274
    BTR-60256195
    BTR-70233196
    D7299274
    T62299273
    T72232196
    ZIL131299274
    ZSU-23_4299274
    Table 1. Data category and number of samples
    AlgorithmNumber of parameters /107Recognition rate in training setRecognition rate in test setTime /s
    ResNet502.599.497.91760
    ResNet1014.590.790.92426
    Inception ResNetV25.588.194.57967
    Table 2. Recognition rates of different algorithms
    Number of parameters /107Recognition rate in training set /%Recognition rate in test set /%Training time /s
    3.299.598.94023
    Table 3. Recognition rate of original residual attention network
    StageNumber of parameters /107Recognition rate in training set /%Recognition rate in test set /%Training time /s
    31.199.598.9
    21.299.599.4
    11.599.799.62850
    03.299.598.94023
    Table 4. Experimental results of different improvement stages
    ModelAverage recognition rate /%
    Model S99.6
    CMNet model[25]99.3
    Faster R-CNN model[26]99.1
    A-ConvNets model98.1
    SVM model[27]90.0
    Table 5. Recognition results of different models
    Noise ratio /%515203035
    Recognition rate /%9999999999
    Table 6. Occlusion recognition results
    Noise ratio /%51015
    Recognition rate /%968882
    Table 7. Recognition results of model under salt and pepper noise
    Baodai Shi, Qin Zhang, Yao Li, Yuhuan Li. SAR Image Target Recognition Based on Improved Residual Attention Network[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0810008
    Download Citation