Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 59, Issue 4, 0410009(2022)

Target Location Algorithm of Contact Network Pipe Cap Based on Improved Faster R-CNN

Guimei Gu1, Chong Chen1、*, Xiaoning Yu1, Cunjun Zhang2, Zhen Tong3, and Xiaoyun Mei1
Author Affiliations
  • 1School of electrical engineering and automation, Lanzhou Jiaotong University, Lanzhou , Gansu 730070, China
  • 2China Railway Lanzhou Bureau Group Co., Ltd., Lanzhou , Gansu 730030, China
  • 3Qingyang Power Supply Company of State Grid Gansu Electric Power Company, Qingyang , Gansu 745000, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (10)
    Equations (16)
    References (18)
    Cited By (1)
    Tools
    Paper Information
    Topics
    Figures & Tables(10)
    Faster R-CNN deep learning model based on VGG16

    Fig. 1. Faster R-CNN deep learning model based on VGG16

    Download full size
    Structure of RPN network

    Fig. 2. Structure of RPN network

    Download full size
    Clustering. (a) Relationship between k value and accuracy; (b) clustering resultwhen k=12

    Fig. 3. Clustering. (a) Relationship between k value and accuracy; (b) clustering resultwhen k=12

    Download full size
    Results of four training losses. (a) VGG16 model; (b) resnet50 model; (c) resnet101 model; (d) resnet152 model

    Fig. 4. Results of four training losses. (a) VGG16 model; (b) resnet50 model; (c) resnet101 model; (d) resnet152 model

    Download full size
    Training loss of different feature extraction network models

    Fig. 5. Training loss of different feature extraction network models

    Download full size
    Relationship between model precision and number of iterations. (a) VGG16 model; (b) resnet50 model; (c) resnet101 model; (d) resnet152 model

    Fig. 6. Relationship between model precision and number of iterations. (a) VGG16 model; (b) resnet50 model; (c) resnet101 model; (d) resnet152 model

    Download full size
    Relationship between model recall and number of iterations. (a) VGG16 model; (b) resnet50 model; (c) resnet101 model; (d) resnet152 model

    Fig. 7. Relationship between model recall and number of iterations. (a) VGG16 model; (b) resnet50 model; (c) resnet101 model; (d) resnet152 model

    Download full size
    Target location results of contact network pipe cap. (a) VGG16 positioning results; (b) K-means+VGG16 positioning results; (c) resnet50 positioning results; (d) K-means+resnet50 positioning results; (e) resnet101 positioning results; (f) K-means+resnet101 positioning results; (g) resnet152 positioning results; (h) K-means+resnet152 positioning results

    Fig. 8. Target location results of contact network pipe cap. (a) VGG16 positioning results; (b) K-means+VGG16 positioning results; (c) resnet50 positioning results; (d) K-means+resnet50 positioning results; (e) resnet101 positioning results; (f) K-means+resnet101 positioning results; (g) resnet152 positioning results; (h) K-means+resnet152 positioning results

    Download full size

    • Table 1. Parameters of VGG 16 network convolution process

      View table

      Table 1. Parameters of VGG 16 network convolution process

      LayerInputConvolution kernelStep lengthOutput
      Conv3-64M×N×33×31M×N×64
      Conv 3-64M×N×643×31M×N×64
      Pool 1M×N×642×22（M/2）×（N/2）×64
      Conv3-128（M/2）×（N/2）×643×31（M/2）×（N/2）×128
      Conv3-128（M/2）×（N/2）×1283×31（M/2）×（N/2）×128
      Pool 2（M/2）×（N/2）×1282×22（M/4）×（N/4）×128
      Conv3-256（M/4）×（N/4）×1283×31（M/4）×（N/4）×256
      Conv3-256（M/4）×（N/4）×2563×31（M/4）×（N/4）×256
      Conv3-256（M/4）×（N/4）×2563×31（M/4）×（N/4）×256
      Pool 3（M/4）×（N/4）×2562×22（M/8）×（N/8）×256
      Conv3-512（M/8）×（N/8）×2563×31（M/8）×（N/8）×512
      Conv3-512（M/8）×（N/8）×5123×31（M/8）×（N/8）×512
      Conv3-512（M/8）×（N/8）×5123×31（M/8）×（N/8）×512
      Pool 4（M/8）×（N/8）×5122×22（M/16）×（N/16）×512
      Conv3-512（M/16）×（N/16）×5123×31（M/16）×（N/16）×512
      Conv3-512（M/16）×（N/16）×5123×31（M 16）×（N/16）×512
      Conv3-512（M/16）×（N/16）×5123×31（M/16）×（N/16）×512

    • Table 2. Comparison of deep learning network

      View table

      Table 2. Comparison of deep learning network

      Extraction networkAccurary /%Recall /%F1 /%Detection time /s
      VGG1672.2975.0073.620.215
      K-means+VGG1681.6088.7485.020.216
      resnet5080.9283.7082.280.283
      K-means+resnet5083.1689.7886.340.283
      resnet10180.9782.6181.780.339
      K-means+resnet10182.3783.7083.020.338
      resnet15275.4979.3577.370.395
      K-means+resnet15282.5184.7883.620.395
    Tools

    Get Citation

    Copy Citation Text

    Guimei Gu, Chong Chen, Xiaoning Yu, Cunjun Zhang, Zhen Tong, Xiaoyun Mei. Target Location Algorithm of Contact Network Pipe Cap Based on Improved Faster R-CNN[J]. Laser & Optoelectronics Progress, 2022, 59(4): 0410009

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Feb. 2, 2021

    Accepted: Mar. 23, 2021

    Published Online: Jan. 25, 2022

    The Author Email: Chong Chen (1334065344@qq.com)

    DOI:10.3788/LOP202259.0410009

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology