• Optics and Precision Engineering
  • Vol. 32, Issue 10, 1552 (2024)
Guanghui LIU1,2,*, Zhe SHAN1,2, Yuanhai YANG1,2, Heng WANG1,2..., Yuebo MENG1,2,* and Shengjun XU1,2|Show fewer author(s)
Author Affiliations
  • 1College of Information and Control Engineering,Xi'an University of Architecture and Technology,Xi'an70055,China
  • 2Xi'an Key Laboratory of Intelligent Technology for Building and Manufacturing,Xi'an710055,China
  • show less
    DOI: 10.37188/OPE.20243210.1552 Cite this Article
    Guanghui LIU, Zhe SHAN, Yuanhai YANG, Heng WANG, Yuebo MENG, Shengjun XU. Optical remote sensing road extraction network based on GCN guided model viewpoint[J]. Optics and Precision Engineering, 2024, 32(10): 1552 Copy Citation Text show less
    Overall structure of the RGGVNet
    Fig. 1. Overall structure of the RGGVNet
    Overall structure of the ConvNeXt
    Fig. 2. Overall structure of the ConvNeXt
    Illustration of the GVPG module
    Fig. 3. Illustration of the GVPG module
    Illustration of the dense guidance viewpoint strategy
    Fig. 4. Illustration of the dense guidance viewpoint strategy
    Overall structure of the decoder
    Fig. 5. Overall structure of the decoder
    Sample images and labels of datasets
    Fig. 6. Sample images and labels of datasets
    Visualized results of different algorithms on the Massachusetts road dataset
    Fig. 7. Visualized results of different algorithms on the Massachusetts road dataset
    Visualized results of different algorithms on the DeepGlobe road dataset
    Fig. 8. Visualized results of different algorithms on the DeepGlobe road dataset
    Test results in wider cities
    Fig. 9. Test results in wider cities
    MethodsPrecision/%Recall/%F1-score/%IoU/%
    FCN69.1670.1469.6553.43
    SegNet75.7678.2676.9961.21
    PSPNet76.3678.1677.2461.37
    U-Net76.1279.0177.5462.52
    LinkNet81.5080.6377.0762.70
    DeeplabV3+74.4078.8576.5661.02
    CDG81.4171.8076.1061.90
    DA-RoadNet79.1677.2578.1961.90
    CADUNet79.4576.5577.8964.19
    DDU-Net82.5473.9978.0363.98
    RGGVNet77.0281.9379.4065.84
    Table 1. Results of comparative experiments on the Massachusetts road dataset
    MethodsPrecisionRecallF1-scoreIoU
    FCN71.3874.7473.0257.51
    SegNet69.4980.0173.2260.43
    PSPNet67.6480.9273.6959.82
    U-Net81.5080.4080.9567.99
    LinkNet77.4579.9878.6864.86
    DeeplabV3+68.7280.4874.1461.02
    D-LinkNet64.12
    RADANet73.6758.58
    BDTNet84.1876.7780.3067.09
    DCS-TransUperNet82.4478.4380.3965.36
    SDUNet78.4074.2079.4066.80
    CoANet81.2268.37
    RENA78.4077.0076.4063.10
    RGGVNet79.5884.3781.9069.36
    Table 2. Results of comparative experiments on the DeepGlobe road dataset
    MethodF1-score/%IoU/%Params.(×106
    Baseline75.8961.5429.663
    Baseline-M76.8762.7930.017
    Baseline-ML77.0163.1230.017
    Baseline-MG78.8865.5533.582
    Baseline-MGS79.1765.7033.582
    Baseline-MGSL79.4065.8433.582
    Table 3. Results of ablation experiment
    αF1-score/%IoU/%
    079.1765.70
    0.279.3365.80
    0.479.4065.84
    0.679.1365.81
    0.878.9765.74
    1.078.6365.48
    Table 4. Experimental results of loss function hyperparametric
    Guanghui LIU, Zhe SHAN, Yuanhai YANG, Heng WANG, Yuebo MENG, Shengjun XU. Optical remote sensing road extraction network based on GCN guided model viewpoint[J]. Optics and Precision Engineering, 2024, 32(10): 1552
    Download Citation