Chinese Journal of Lasers, Volume. 45, Issue 3, 307013(2018)

Fluorescent Microsphere Segmentation and Classification Based on Watershed and Semi-Supervised Minor Reconstruction Error

Huang Hong, Jin Yingying, Li Zhengying, Duan Yule, and Shi Guangyao
Author Affiliations
  • [in Chinese]
  • show less
    Figures & Tables(7)
    Flowchart of the fluorescent microspheres classification algorithm
    Grayscale histogram of the fluorescent microspheres image
    Results with improved watershed segmentation. (a1)(a2) Original images of the fluorescent microspheres; (b1)(b2) grayscale maps after semi-threshold foreground extraction; (c1)(c2) segmented results without improvement; (d1)(d2) gradient images; (e) final segmented results
    Statistical distribution of fluorescent microsphere objects in HSV color space. Statistical distributions of (a) hue channel; (b) saturation channel; (c) value channel
    Seven different types of fluorescent microsphere samples classes. (a) Class 1, (b) class 2, (c) class 3, (d) class 4, (e) class 5, (f) class 6, and (g) class 7
    • Table 1. Overall classification accuracies and run times of different methods with different labeled data samples per class

      View table

      Table 1. Overall classification accuracies and run times of different methods with different labeled data samples per class

      Classificationmethod2 labels4 labels6 labels8 labels
      Overallclassificationaccuracy /%Runtime /sOverallclassificationaccuracy /%Runtime /sOverallclassificationaccuracy /%Runtime /sOverallclassificationaccuracy /%Runtime /s
      LDA70.0±0.0660.08776.4±0.0440.12379.4±0.0220.13581.1±0.0270.135
      KNN77.2±0.0640.00481.0±0.0390.00582.0±0.0230.00682.4±0.0280.007
      RF77.3±0.0443.80180.3±0.0223.90481.7±0.0194.02082.7±0.0134.028
      SRC72.9±0.0501.28177.9±0.0341.32680.1±0.0241.36081.1±0.0331.375
      SVM76.1±0.0501.14680.0±0.0251.25481.7±0.0281.40283.0±0.0161.557
      SSMREC84.3±0.0305.91985.7±0.0195.95486.2±0.0216.03386.5±0.0236.093
    • Table 2. Each class classification accuracies and overall accuracies in different methods with 2 labeled data samples per class%

      View table

      Table 2. Each class classification accuracies and overall accuracies in different methods with 2 labeled data samples per class%

      ClassificationmethodClassification accuracy of each classOverallaccuracy
      Class 1Class 2Class 3Class 4Class 5Class 6Class 7
      LDA73.084.769.265.356.873.469.069.8
      KNN77.292.485.965.066.179.677.177.2
      RF92.992.573.362.355.972.590.177.3
      SRC69.691.185.866.056.879.867.572.9
      SVM74.891.285.764.068.779.274.476.2
      SSMREC95.595.386.366.669.580.393.384.3
    Tools

    Get Citation

    Copy Citation Text

    Huang Hong, Jin Yingying, Li Zhengying, Duan Yule, Shi Guangyao. Fluorescent Microsphere Segmentation and Classification Based on Watershed and Semi-Supervised Minor Reconstruction Error[J]. Chinese Journal of Lasers, 2018, 45(3): 307013

    Download Citation

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

    Special Issue:

    Received: Sep. 5, 2017

    Accepted: --

    Published Online: Mar. 6, 2018

    The Author Email:

    DOI:10.3788/CJL201845.0307013

    Topics