Journal of Applied Optics, Volume. 46, Issue 3, 612(2025)

Semantic segmentation of remote sensing images based on multi-scale feature interaction and boundary optimization

Han ZHANG1, Alex Hay-Man NG2、*, and Xun LIU1
Author Affiliations
  • 1School of Information Engineering, Guangdong University of Technology, Guangzhou 510006, China
  • 2School of Civil and Transportation Engineering, Guangdong University of Technology, Guangzhou 510006, China
  • show less
    Figures & Tables(10)
    Framework diagram of overall structure
    Structure diagram of multi-scale feature interaction
    Structure diagram of boundary optimization
    Example of training set for remote sensing images
    Comparison of segmentation results
    Comparison of details
    Experimental results
    Loss curves
    • Table 1. Semantic information of labels

      View table
      View in Article

      Table 1. Semantic information of labels

      ClassNo.Label color ( R, G, B )
      Built-up0(255,0,0)
      Forest1(0,255,255)
      Farmland2(0,255,0)
      Meadow3(255,255,0)
      Water4(0,0,255)
      Others5(0,0,0)
    • Table 2. Evaluation indicators for each structure

      View table
      View in Article

      Table 2. Evaluation indicators for each structure

      StructureEvaluationBuilt-upWaterForestMeadowFarmland
      U-NetIoU0.60320.81650.62690.71480.6327
      Accuracy0.66740.83580.80520.77150.7063
      Precision0.82360.83910.70220.78350.8164
      F-Score0.67370.84890.73460.78680.7441
      U-Net-HEDIoU0.62490.83080.64830.73050.6591
      Accuracy0.69060.85760.82230.78320.7188
      Precision0.85130.87660.73170.80740.8519
      F-Score0.69640.87490.75970.80440.7684
      Mask2FormerIoU0.64110.89830.68550.76940.6939
      Accuracy0.71000.92510.88180.85560.7446
      Precision0.93800.96820.79200.88200.9031
      F-Score0.77870.94710.79430.85620.8093
      Mask2Former-MS (Ours)IoU0.67440.92400.70690.79530.7183
      Accuracy0.74150.94440.90390.89790.8790
      Precision0.95780,98860.81110.91110.9253
      F-Score0.80840.96050.82140.88600.8365
      Mask2Former-BR(Ours)IoU0.67870.93300.71360.79670.7202
      Accuracy0.74260.94530.94110.89340.8809
      Precision0.96350.98290.83990.92450.9255
      F-Score0.81140.95990.81640.87430.8373
    Tools

    Get Citation

    Copy Citation Text

    Han ZHANG, Alex Hay-Man NG, Xun LIU. Semantic segmentation of remote sensing images based on multi-scale feature interaction and boundary optimization[J]. Journal of Applied Optics, 2025, 46(3): 612

    Download Citation

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

    Category:

    Received: Jul. 5, 2024

    Accepted: --

    Published Online: May. 28, 2025

    The Author Email: Alex Hay-Man NG (吴希文)

    DOI:10.5768/JAO202546.0302001

    Topics