Spectroscopy and Spectral Analysis, Volume. 42, Issue 9, 2862(2022)

Multispectral Lightweight Ship Target Detection Algorithm for Sentinel-2 Satellite

Li CHEN1; 2; 3;, Shi-yong WANG1; 3;, Si-li GAO1; 3;, Chang TAN1; 2; 3;, and Lin-han LI1; 2; 3;
Author Affiliations
  • 1. Shanghai Institute of Technical Physics, Chinese Academy of Sciences, Shanghai 200083, China
  • show less
    Figures & Tables(16)
    Comparison of images before and after atmospheric correction(a): Before atmospheric correction; (b): After atmospheric correction
    Algorithm flow chart
    Spectral curves of typical surface features
    Comparison of NDWI and PNDWI extractions for seawater with and without sediment(a): RGB image of seawater with sediment; (b): NIR image of seawater with sediment; (c): NDWI for seawater with sediment; (d): PNDWI for seawater with sediment; (e): RGB image of seawater without sediment; (f): NIR image of seawater without sediment; (g): NDWI for seawater without sediment; (h): PNDWI for seawater without sediment
    Statistical box plots of pixel values of 5 scenes in 4 bands and box plot interpretation(a): Ship; (b): Thick clouds; (c): Thin clouds; (d): Smooth sea; (e): Sea clutter; (f): Box plot interpretation
    Process of assisting determination(a): RGB; (b): NIR; (c): PNDWI; (d): Assisting determination
    An example of candidate region slice extraction
    (a)Schematic diagram of spectral feature extraction, (b)1×1 convolution diagram
    Channel shuffling diagram
    1×1 Diagram of the LSGFNet network structure
    Part of the sample data for fine detection
    Loss reduction (a) and precision variation (b) of verification set with an interation of 10 000
    Detection result(a): RGB; (b): NIR; (c): Binary images of rough detection; (d): Fine detection
    • Table 1. Rough dectection results of ships

      View table
      View in Article

      Table 1. Rough dectection results of ships

      Recall
      粗检测98.99%
    • Table 2. Comparison of lightweight network algorithms for test set

      View table
      View in Article

      Table 2. Comparison of lightweight network algorithms for test set

      Precision/%FLOPs/MParams/M
      ResNet1895.86557.2411
      MobileNetV294.8794.892.3
      ShuffleNetV295.2246.341.25
      LSGFNet96.0438.991.1
    • Table 3. Comparison of detection results by different algorithms

      View table
      View in Article

      Table 3. Comparison of detection results by different algorithms

      YoloV3SSDProposed
      AP90.10%88.49%92.98%
    Tools

    Get Citation

    Copy Citation Text

    Li CHEN, Shi-yong WANG, Si-li GAO, Chang TAN, Lin-han LI. Multispectral Lightweight Ship Target Detection Algorithm for Sentinel-2 Satellite[J]. Spectroscopy and Spectral Analysis, 2022, 42(9): 2862

    Download Citation

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

    Category: Research Articles

    Received: Aug. 19, 2021

    Accepted: Nov. 23, 2021

    Published Online: Nov. 17, 2022

    The Author Email:

    DOI:10.3964/j.issn.1000-0593(2022)09-2862-08

    Topics