Remote Sensing Technology and Application, Volume. 40, Issue 3, 708(2025)

Multi-source Scene Feature Transformer Fusion Method for Urban Functional Area Identification

Zhiwei XIE1, Lei HAN1, Lishuang SUN1、*, and Bo PENG2
Author Affiliations
  • 1School of Transportation and Geomatics Engineering, Shenyang Jianzhu University, Shenyang110168, China
  • 2School of Geomatics Engineering, Liaoning Vocational College of Ecological Engineering, Shenyang110112, China
  • show less
    Figures & Tables(19)
    Study area
    Data source
    Transformer-based RP framework
    ResNet-50 framework
    TAZ construction and manual labeling
    The identification results of each model
    The first local identification results of each model
    The second local identification results of each model
    GCN parameter experiment
    The identification results of ablation experiment
    The first local identification results of ablation experiment
    The second local identification results of ablation experiment
    • Table 1. POI data categories

      View table
      View in Article

      Table 1. POI data categories

      序号POI类别序号POI类别
      1生活服务8住宅区
      2公共设施9商务住宅相关
      3交通设施10住宿服务
      4科教文化11公司企业
      5体育休闲12餐饮服务
      6医疗保健13购物服务
      7政府社会团体14金融保险
    • Table 2. POI reclassification

      View table
      View in Article

      Table 2. POI reclassification

      一级分类二级分类三级分类
      公共

      政府机构、医疗、科教

      文化等

      政府机构、医院、图书馆、火

      车站等

      居住住宅区、商务住宅相关住宅小区、别墅、社区中心等
      商业餐饮、购物、住宿、金融中餐厅、酒店、电影院、银行等
      工业公司企业、商务住宅相关公司、工厂、科技园、产业园等
    • Table 3. Overall accuracy and Kappa coefficient of identification for each model

      View table
      View in Article

      Table 3. Overall accuracy and Kappa coefficient of identification for each model

      方法数据测试集
      影像POI总体精度Kappa系数
      ResNet-500.650.36
      ResNet-500.700.51
      GCN0.710.52
      GCN0.750.60
      Transformer0.820.70
    • Table 4. ResNet-50 parameter experiment

      View table
      View in Article

      Table 4. ResNet-50 parameter experiment

      Batch_size总体精度Kappa系数
      80.600.28
      160.650.36
      320.590.28
      640.570.26
    • Table 5. The parameter experiment of transformer

      View table
      View in Article

      Table 5. The parameter experiment of transformer

      层数总体精度Kappa系数
      10.810.70
      20.760.64
      30.740.61
      40.820.70
      50.800.69
      60.740.62
    • Table 6. The accuracy statistics of ablation experiments

      View table
      View in Article

      Table 6. The accuracy statistics of ablation experiments

      方法自然场景特征社会场景特征OAKappa
      a0.700.41
      b0.800.66
      c0.820.70
    • Table 7. The parameter number and calculation cost of each model

      View table
      View in Article

      Table 7. The parameter number and calculation cost of each model

      方法参数训练时间Epoch/s
      Resnet-5021 797 6720.15
      GCN91 3360.09
      Transformer9 590 7840.24
    Tools

    Get Citation

    Copy Citation Text

    Zhiwei XIE, Lei HAN, Lishuang SUN, Bo PENG. Multi-source Scene Feature Transformer Fusion Method for Urban Functional Area Identification[J]. Remote Sensing Technology and Application, 2025, 40(3): 708

    Download Citation

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

    Category:

    Received: Dec. 19, 2023

    Accepted: --

    Published Online: Sep. 28, 2025

    The Author Email: Lishuang SUN (sunlishuang@sjzu.edu.cn)

    DOI:10.11873/j.issn.1004-0323.2025.3.0708

    Topics