Laser & Optoelectronics Progress, Volume. 55, Issue 5, 051010(2018)

Surface Crack Detection Algorithm Based on Double Threshold and Tensor Voting

Huixian Li, Bin Zhang*, Dan Liu, Tengda Yang, Wenhao Song, and Fengyu Li
Author Affiliations
  • College of Physical Engineering, Zhengzhou University, Zhengzhou, Henan 450001, China
  • show less
    Figures & Tables(10)
    Acquisition of the ROI image. (a) Mask process schematic; (b) original image; (c) mask image; (d) ROI image
    Flow chart of double threshold algorithm
    Bilateral filtering image after increase the dimension
    Tensor voting. (a) Ball voting field; (b) stick voting field; (c) schematic of tensor voting rules; (d) probability significance map
    Complete crack acquisition process
    (a) Experimental equipment; (b) image of components
    Performance evaluation of algorithm. (a) Different threshold; (b) different σ values; (c) runtime
    Image processing and results with double threshold. (a) ROI images; (b) images obtained by Otsu algorithm; (c) images obtained by the double threshold
    Crack detection on five images. (a) Original images; (b) detection results by crack connection algorithm; (c) cracks detection results by proposed algorithm; (d) truth cracks
    • Table 1. Three evaluation parameters of two algorithms at the best F value

      View table

      Table 1. Three evaluation parameters of two algorithms at the best F value

      IndexCrack fragments' connection algorithmAlgorithm based on double threshold and tensor voting
      Precision0.580.81
      Recall0.750.82
      F0.640.81
    Tools

    Get Citation

    Copy Citation Text

    Huixian Li, Bin Zhang, Dan Liu, Tengda Yang, Wenhao Song, Fengyu Li. Surface Crack Detection Algorithm Based on Double Threshold and Tensor Voting[J]. Laser & Optoelectronics Progress, 2018, 55(5): 051010

    Download Citation

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

    Category: Image processing

    Received: Oct. 16, 2017

    Accepted: --

    Published Online: Sep. 11, 2018

    The Author Email: Zhang Bin (13503811569@163.com)

    DOI:10.3788/LOP55.051010

    Topics