• Acta Photonica Sinica
  • Vol. 51, Issue 2, 0210008 (2022)
Yanni WANG, Xuesong SUN*, and Lixian YU
Author Affiliations
  • School of Information and Control Engineering,Xi′an University of Architecture and Technology,Xi′an710055,China
  • show less
    DOI: 10.3788/gzxb20225102.0210008 Cite this Article
    Yanni WANG, Xuesong SUN, Lixian YU. Lightweight Synthetic Aperture Radar Ship Detection Algorithm with Enhanced Receptive Field[J]. Acta Photonica Sinica, 2022, 51(2): 0210008 Copy Citation Text show less
    The proposed network structure
    Fig. 1. The proposed network structure
    The operation process of channel shuffle
    Fig. 2. The operation process of channel shuffle
    Building blocks of ShuffleNetV2
    Fig. 3. Building blocks of ShuffleNetV2
    Spatial Attention Module
    Fig. 4. Spatial Attention Module
    SAM heatmap
    Fig. 5. SAM heatmap
    The structure of SSPP
    Fig. 6. The structure of SSPP
    Feature map of SSPP
    Fig. 7. Feature map of SSPP
    PANet structure
    Fig. 8. PANet structure
    Improved PANet structure
    Fig. 9. Improved PANet structure
    Sample images and labels in SSDD
    Fig. 10. Sample images and labels in SSDD
    Mosaic data enhancement
    Fig. 11. Mosaic data enhancement
    Target detection results of the improved algorithm in different scenarios
    Fig. 12. Target detection results of the improved algorithm in different scenarios
    P-R curve(the threshold changed from 0.05 to 0.95)
    Fig. 13. P-R curve(the threshold changed from 0.05 to 0.95)
    Comparison of detection algorithms for complex background of near-shore ships
    Fig. 14. Comparison of detection algorithms for complex background of near-shore ships
    Comparison of detection algorithms for small target ships in the open sea
    Fig. 15. Comparison of detection algorithms for small target ships in the open sea
    LayerKernel sizeStrideRepeatOutput sizeOutput channels(1×)
    Image---224×2243

    Conv1

    MaxPool

    3×3

    3×3

    2

    2

    1

    1

    112×112

    56×56

    24

    24

    Stage2

    -

    -

    2

    1

    1

    3

    28×28

    28×28

    116
    Stage3

    -

    -

    2

    1

    1

    7

    14×14

    14×14

    232
    Stage4

    -

    -

    2

    1

    1

    3

    7×7

    7×7

    464
    Conv51×1117×71 024
    GlobalPool7×7--7×7-
    FC----1 000
    FLOPs----146 M
    Weights----2.3 M
    Table 1. Overall architecture of ShuffleNetV2
    SSPPSAMPANmAP/%Speed/FPS
    ×××85.652
    ××90.250
    ×91.548
    ×93.447
    94.746
    Table 2. Accuracy comparison results of different modules
    AlgorithmSSDYOLOv3YOLOv4YOLOv3-tinyYOLOv4-tinyOurs
    Near shorePrecision/%74.275.378.469.372.380.6
    Recall/%91.692.491.578.583.894.3
    Open seaPrecision/%87.392.396.787.591.297.8
    Recall/%93.495.498.283.485.697.2
    Table 3. Comparison of different detection algorithms in near shore and open sea scenes
    AlgorithmBackboneFLOPsParamsSpeed/FPSModel size/MBmAP/%
    SSDVGG16130.9 B99.6 M2810089.3
    YOLOv3Darknet53156.3 B246.1 M2623692.3
    YOLOv4CSPDarknet53128.5 B255.7 M2724495.6
    YOLOv3-tinyDarknet53_tiny5.62 B8.86 M3533.490.3
    YOLOv4-tinyCSPdarknet53_tiny6.96 B6.06 M3822.693.8
    OursShuffleNetV21.52 B2.34 M465.394.7
    Table 4. Performance comparison of different detection algorithms
    Yanni WANG, Xuesong SUN, Lixian YU. Lightweight Synthetic Aperture Radar Ship Detection Algorithm with Enhanced Receptive Field[J]. Acta Photonica Sinica, 2022, 51(2): 0210008
    Download Citation