• Laser & Optoelectronics Progress
  • Vol. 58, Issue 22, 2210002 (2021)
Ke Zhang and Liang Zhang*
Author Affiliations
  • College of Electronic Information and Automation, Civil Aviation University of China, Tianjin 300300, China
  • show less
    DOI: 10.3788/LOP202158.2210002 Cite this Article Set citation alerts
    Ke Zhang, Liang Zhang. Multi-Scale Detection for X-Ray Prohibited Items in Complex Background[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2210002 Copy Citation Text show less
    Framework of M2Det
    Fig. 1. Framework of M2Det
    Structure of different modules in FFM. (a)FFMv1; (b)FFMv2
    Fig. 2. Structure of different modules in FFM. (a)FFMv1; (b)FFMv2
    Structure of different models. (a) FPN; (b) RFP
    Fig. 3. Structure of different models. (a) FPN; (b) RFP
    Structure of different modules. (a) CAM; (b) SAM
    Fig. 4. Structure of different modules. (a) CAM; (b) SAM
    EM2Det model
    Fig. 5. EM2Det model
    Structure of feature fusion enhancement module
    Fig. 6. Structure of feature fusion enhancement module
    Detail of feature fusion enhancement module
    Fig. 7. Detail of feature fusion enhancement module
    Details of recursive modules for U-shaped network
    Fig. 8. Details of recursive modules for U-shaped network
    Residual side attention module
    Fig. 9. Residual side attention module
    Some images in SIXray OD dataset. (a) Gun; (b) wrench; (c) knife; (d) pliers; (e) scissors
    Fig. 10. Some images in SIXray OD dataset. (a) Gun; (b) wrench; (c) knife; (d) pliers; (e) scissors
    Visualizations in different situations. (a) Scissors; (b) pliers; (c) gun; (d) wrench; (e) knife
    Fig. 11. Visualizations in different situations. (a) Scissors; (b) pliers; (c) gun; (d) wrench; (e) knife
    Number of TUMsParameters /MbitmAP /%
    2182.0573.34
    4227.8873.99
    6293.6176.89
    8331.6279.03
    10383.7879.62
    Table 1. Index under different number of TUMs
    StructureIntegration of pool 5mAP /%
    Baseline(conv 4_3+pool 5)×79.03(0)
    Three layers (pool 3+conv 4_3+pool 5)×81.94(2.91↑)
    Three layers (pool 3+conv 4_3+pool 5)82.82(3.79↑)
    Four layers (pool 2+pool 3+conv 4_3+pool 5)×81.13(2.10↑)
    Four layers (pool 2+pool 3+conv 4_3+pool 5)82.05(3.02↑)
    Table 2. Experiment effect of feature fusion enhancement module
    StructureParameters /MbitResidual edge attention modulemAP /%
    8 TUM334.50×82.82
    1 RTUM +7 TUM356.67×83.61(0.79↑)
    8 RTUM514.89×84.09(1.27↑)
    8 RTUM+CBAM517.89(3.00↑)×85.12(2.30↑)
    85.40(2.58↑)
    8 RTUM+ECA511.86(3.00↓)×84.86(2.04↑)
    85.02(2.20↑)
    Table 3. Experimental effects of U-shaped network recursive module and residual edge attention module
    FrameworkAccuracy /%mAP /%
    GunPilersScissorsKnifeWrench
    SSD927266555270.9
    Faster RCNN937571656774.2
    YOLOv3927164535166.2
    M2Det948376746879.0
    M2Det+EFFM968682777382.8
    M2Det+EFFM+RTUM968882787684.0
    EM2Det988984797785.4
    Table 4. Accuracy of different detection models
    Ke Zhang, Liang Zhang. Multi-Scale Detection for X-Ray Prohibited Items in Complex Background[J]. Laser & Optoelectronics Progress, 2021, 58(22): 2210002
    Download Citation