Laser & Optoelectronics Progress, Volume. 60, Issue 16, 1628004(2023)

Road Extraction from Remote Sensing Image Based on an Improved U-Net

Zhe He1,2, Yuxiang Tao1,2、*, Xiaobo Luo1,2, and Hao Xu1,2
Author Affiliations
  • 1School of Computer Sciences and Technology, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
  • 2Spatial Big Data Research Center, Chongqing University of Posts and Telecommunications, Chongqing 400065, China
  • show less
    Figures & Tables(9)
    Overall structure of HSA-UNet
    Comparison of three structures. (a) Plain unit; (b) residual unit; (c) ARLU
    AFF module
    AE-ASPP module
    SAM
    Training samples and labels
    Road prediction diagrams of proposed network and other networks
    • Table 1. Comparison of evaluation result of ablation experiment

      View table

      Table 1. Comparison of evaluation result of ablation experiment

      No.BASEARLUAE-ASPPRIoU /%
      162.41
      264.65

      3

      4

      64.24

      65.60

    • Table 2. Comparison of road extraction result of different networks unit: %

      View table

      Table 2. Comparison of road extraction result of different networks unit: %

      ModelRPreciousRRecallSF1RIoU
      U-Net80.7173.3476.8162.41
      SegNet75.8274.0974.9559.93
      ResUnet81.4272.5376.7262.23
      DeepLabV3+80.0170.4674.9459.92
      D-LinkNet75.3579.6777.4563.20
      NL-LinkNet75.2780.2677.6763.51
      GC-DCNN82.6474.2878.2464.26
      HSA-UNet80.5077.9979.2365.60
    Tools

    Get Citation

    Copy Citation Text

    Zhe He, Yuxiang Tao, Xiaobo Luo, Hao Xu. Road Extraction from Remote Sensing Image Based on an Improved U-Net[J]. Laser & Optoelectronics Progress, 2023, 60(16): 1628004

    Download Citation

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

    Category: Remote Sensing and Sensors

    Received: Sep. 26, 2022

    Accepted: Nov. 24, 2022

    Published Online: Aug. 18, 2023

    The Author Email: Tao Yuxiang (taoyx@cqupt.edu.cn)

    DOI:10.3788/LOP222634

    Topics