• Laser & Optoelectronics Progress
  • Vol. 58, Issue 12, 1210022 (2021)
Quan Wang1 and Benshun Yi2、*
Author Affiliations
  • 1FiberHome Technologies Group, Wuhan, Hubei 430074, China
  • 2Wuhan University, Wuhan, Hubei 430072, China
  • show less
    DOI: 10.3788/LOP202158.1210022 Cite this Article Set citation alerts
    Quan Wang, Benshun Yi. Insulator Defect Recognition in Aerial Images Based on Gaussian YOLOv3[J]. Laser & Optoelectronics Progress, 2021, 58(12): 1210022 Copy Citation Text show less
    Image of the insulator. (a) Normal image; (b) image with defects
    Fig. 1. Image of the insulator. (a) Normal image; (b) image with defects
    Structure of the YOLOv3
    Fig. 2. Structure of the YOLOv3
    Test results of the YOLOv3. (a) Target is not completely wrapped; (b) duplicate frame
    Fig. 3. Test results of the YOLOv3. (a) Target is not completely wrapped; (b) duplicate frame
    Output of the YOLOv3
    Fig. 4. Output of the YOLOv3
    Output of the Gaussian YOLOv3
    Fig. 5. Output of the Gaussian YOLOv3
    Training losses in different situations. (a) Step 1); (b) step 2); (c) step 3)
    Fig. 6. Training losses in different situations. (a) Step 1); (b) step 2); (c) step 3)
    Detection results of the Gaussian YOLOv3. (a) Normal insulator; (b) insulator with defects
    Fig. 7. Detection results of the Gaussian YOLOv3. (a) Normal insulator; (b) insulator with defects
    Data setInsulatorDefect
    Training set1727713
    Validation set19379
    Test set480200
    Table 1. Division of the data set
    AlgorithmInsulator /%Defect /%mAP /%XFPS /(frame·s-1)Pi /%Pd /%
    Faseter R-CNN91.583.087.15.432.522.0
    YOLOv393.585.090.326.329.624.0
    Guassian YOLOv393.894.594.326.191.794.0
    Table 2. Test results of different algorithms
    Quan Wang, Benshun Yi. Insulator Defect Recognition in Aerial Images Based on Gaussian YOLOv3[J]. Laser & Optoelectronics Progress, 2021, 58(12): 1210022
    Download Citation