Chinese Journal of Lasers, Volume. 46, Issue 8, 0804002(2019)

Automatic Extraction and Classification of Road Markings Based on Deep Learning

Gang Huang1,2、* and Xianlin Liu3
Author Affiliations
  • 1 College of Resource Environment and Tourism, Capital Normal University, Beijing 100048, China
  • 2 Beijing GEO-Vision Tech. Co., Ltd., Beijing 100070, China
  • 3 Chinese Academy of Surveying & Mapping, Beijing 100830, China;
  • show less
    Figures & Tables(8)
    Vehicle-borne mobile measurement system
    Flow chart of automatic extraction and classification for road markings
    Structure of Deeplab V3+
    Sample production. (a) Intensity characteristic image; (b) labeled image
    Validation results
    Road marking vectorization. (a) Classification results; (b) clustering; (c) vectorization
    • Table 1. Accuracy of automatic extraction and classification

      View table

      Table 1. Accuracy of automatic extraction and classification

      ClassificationAccuracy /%
      Mesh88.17
      Sidewalk94.34
      Entrance86.51
      Arrow90.83
      Deceleration vertical81.47
      Deceleration horizontal76.36
      Mark77.62
      Longsolid95.73
      Dotted91.67
      Transverse92.35
      MIoU85.47
    • Table 2. Comparison of different methods

      View table

      Table 2. Comparison of different methods

      MethodRPrecision /%RCompleteness /%Fscore /%
      Ref.[2]77.0353.7963.34
      Ref.[13]93.2773.8682.44
      Ref.[23]74.2966.5370.20
      Proposed92.5987.8490.15
    Tools

    Get Citation

    Copy Citation Text

    Gang Huang, Xianlin Liu. Automatic Extraction and Classification of Road Markings Based on Deep Learning[J]. Chinese Journal of Lasers, 2019, 46(8): 0804002

    Download Citation

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

    Category: measurement and metrology

    Received: Feb. 26, 2019

    Accepted: Apr. 2, 2019

    Published Online: Aug. 13, 2019

    The Author Email: Huang Gang (hgminisar@163.com)

    DOI:10.3788/CJL201946.0804002

    Topics