Spacecraft Recovery & Remote Sensing, Volume. 45, Issue 6, 151(2024)

High-Resolution Remote Sensing Image Building Extraction Method Based on MFF-DeeplabV3+

Siyan LIU1...2, Chunyue WANG1,2, Lu FU1,2, and Ling LI12 |Show fewer author(s)
Author Affiliations
  • 1Chang Guang Satellite Technology Co. Ltd., Changchun 130000, China
  • 2Key Laboratory of Satellite Remote Sensing Application Technology of Jilin Province, Changchun 130000, China
  • show less
    Figures & Tables(13)
    [in Chinese]
    [in Chinese]
    [in Chinese]
    [in Chinese]
    [in Chinese]
    [in Chinese]
    [in Chinese]
    [in Chinese]
    • Table 1. Hardware and software environment configuration information

      View table
      View in Article

      Table 1. Hardware and software environment configuration information

      软(硬)件名称CPU内存大小显卡操作系统CUDAPythonPytorch开发环境
      参数信息Intel Xeon Gold 6254128 GBGeForce RTX 3090Windows Server 201911.23.7.111.7.1Visual Studio Code 1.70.1
    • Table 2. Model parameter configuration

      View table
      View in Article

      Table 2. Model parameter configuration

      参数名称批处理大小最大迭代次数初始学习率优化器
      参数值81500.0001Adam
    • Table 3. Experimental indicators for different backbone networks %

      View table
      View in Article

      Table 3. Experimental indicators for different backbone networks %

      主干网络PRF1IoU
      ResNet10192.8192.2092.5086.06
      SE_ResNet10194.0793.7093.8888.48
      ResNeXt101_32x4d93.7692.4%93.1187.10
      SE_ResNeXt101_32x4d94.0093.4793.7388.20
    • Table 4. MFF experimental indicators %

      View table
      View in Article

      Table 4. MFF experimental indicators %

      主干网络PRF1IoU
      Efficientnet-b092.9690.8491.8985.00
      Efficientnet-b0-MFF92.5491.4091.9785.13
      MobileNetV291.8491.8891.8684.95
      MobileNetV2-MFF92.7890.9591.8684.94
      timm-gernet_l92.7391.3692.0485.25
      timm-gernet_l-MFF93.2691.2092.2285.56
      timm-skResNeXt50_32x4d93.8192.6593.2387.32
      timm-skResNeXt50_32x4d-MFF93.2792.8093.2687.37
      SE_ResNet15293.3994.0794.0088.68
      SE_ResNet152-MFF94.1893.9794.0888.82
      SENet15494.6693.3093.9888.64
      SENet154-MFF94.8294.2494.5389.63
    • Table 5. Experimental indicators for different methods %

      View table
      View in Article

      Table 5. Experimental indicators for different methods %

      方法名称PRF1IoU
      UNet-3Plus94.0393.3093.6688.09
      MobileNetV394.2192.6093.4087.62
      SegNet93.6392.7793.2087.27
      PSPNet92.8091.8292.3185.72
      EncNet94.1893.1493.6688.07
      CDCA-DLV3+94.3492.8293.5787.93
      MANet93.0392.3192.6786.35
      HRNetV2_W4892.1093.7392.9186.75
      本文方法94.8294.2494.5389.63
    Tools

    Get Citation

    Copy Citation Text

    Siyan LIU, Chunyue WANG, Lu FU, Ling LI. High-Resolution Remote Sensing Image Building Extraction Method Based on MFF-DeeplabV3+[J]. Spacecraft Recovery & Remote Sensing, 2024, 45(6): 151

    Download Citation

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

    Category:

    Received: May. 17, 2024

    Accepted: --

    Published Online: Jan. 23, 2025

    The Author Email:

    DOI:10.3969/j.issn.1009-8518.2024.06.013

    Topics