• Laser & Optoelectronics Progress
  • Vol. 58, Issue 8, 0815003 (2021)
Zhihao Chen1、2、*, Yewei Xiao1、2、**, Zhiqiang Li1, and Yang Liu1
Author Affiliations
  • 1School of Automation and Electronic Information, Xiangtan University, Xiangtan, Hunan 411105, China
  • 2Key Laboratory of Intelligent Computing & Information Processing of Ministry of Education, Xiangtan University, Xiangtan, Hunan 411105, China;
  • show less
    DOI: 10.3788/LOP202158.0815003 Cite this Article Set citation alerts
    Zhihao Chen, Yewei Xiao, Zhiqiang Li, Yang Liu. Insulators Identification for Overhead Transmission Lines in Distribution Networks Based on Multi-Scale Dense Network[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0815003 Copy Citation Text show less
    Insulator detection based on the YOLOv3 network
    Fig. 1. Insulator detection based on the YOLOv3 network
    Structure of the Darknet-53
    Fig. 2. Structure of the Darknet-53
    Structure of the DenseNet
    Fig. 3. Structure of the DenseNet
    Structure of the densely connected network
    Fig. 4. Structure of the densely connected network
    Structure of the SPP module
    Fig. 5. Structure of the SPP module
    Structuref the DSM-Darknet
    Fig. 6. Structuref the DSM-Darknet
    Insulator data set. (a) GS; (b) PS; (c) PPin; (d) PPost; (e) PC; (f) R
    Fig. 7. Insulator data set. (a) GS; (b) PS; (c) PPin; (d) PPost; (e) PC; (f) R
    Results of the K-means clustering
    Fig. 8. Results of the K-means clustering
    Training loss curve of the insulator data set
    Fig. 9. Training loss curve of the insulator data set
    PR curves of different insulators
    Fig. 10. PR curves of different insulators
    ROC of different insulators
    Fig. 11. ROC of different insulators
    mAP of different algorithms
    Fig. 12. mAP of different algorithms
    Detection effect of the multi-scale dense network
    Fig. 13. Detection effect of the multi-scale dense network
    CategoryTotalTrainTrainValTest
    GS508530517621272
    PS8872532313302219
    PPin22451347336562
    PPost595635738931490
    PC6839410310251711
    R474528477111187
    Table 1. Number of different categories in the data set
    AlgorithmAnchorInput sizeDense NetSPPMulti-scaleLossGS /%PS /%PPin /%PPost /%PC /%R /%mAP /%FPS /frame
    YOLOv388.689.173.574.289.174.281.540
    Improve189.290.675.376.590.376.683.141
    Improve289.691.176.978.191.178.184.240
    Improve391.892.882.981.893.981.587.534
    Improve493.194.284.885.795.085.789.833
    Improve594.395.786.487.397.388.291.531
    Ours96.196.688.589.898.190.993.428
    Table 2. Different strategies influence the results of the algorithm
    AlgorithmBackboneIterationsPrecision /%Recall /%mAP /%FPS /frame
    Faster R-CNNVGG16-Net6000079.177.574.67
    SSDVGG16-Net6000082.083.278.926
    YOLOv3DarkNet536000083.783.581.540
    RetinaNetResNet1016000084.884.382.420
    OursDSM-DarkNet6000094.596.093.428
    Table 3. Test results of different algorithms
    Zhihao Chen, Yewei Xiao, Zhiqiang Li, Yang Liu. Insulators Identification for Overhead Transmission Lines in Distribution Networks Based on Multi-Scale Dense Network[J]. Laser & Optoelectronics Progress, 2021, 58(8): 0815003
    Download Citation