Fig 1.
Overall framework of the proposed fine-scale urban land cover change assessment method.
All elements were created by the authors.
Fig 2.
Location map of the study area.
(A) the Langfang dataset (China). (B) the Potsdam dataset (Germany). (C) the Guiyang dataset (China). The global location map was generated using the Natural Earth dataset, which is in the public domain. Representative remote sensing imagery was obtained from the Copernicus Data Space Ecosystem (Copernicus Sentinel data), the ISPRS Potsdam dataset and processed by the authors. All map elements and annotations were created by the authors.
Fig 3.
Example image patches and corresponding ground truth labels from the Potsdam dataset.
The figure was generated by the authors for visualization purposes.
Table 1.
Evaluation index.
Fig 4.
Visual comparison of urban land cover classification results produced by different models on the the Potsdam dataset.
All classification results were generated by the authors.
Table 2.
Quantitative performance comparison of different semantic segmentation models on three urban remote sensing datasets.
Fig 5.
Multi-temporal urban land cover classification results for the Langfang Economic Development Zone.
The background remote sensing imagery was obtained from the Copernicus Data Space Ecosystem (Copernicus Sentinel data) and processed by the authors. The imagery represents original satellite observations rather than an online basemap. All classification maps, boundaries, annotations, and map elements were generated by the authors.
Fig 6.
Comparative charts of multiple indicators across the 34 subregions of the Langfang Economic Development Zone (2017–2023).
All map elements were generated by the authors.
Fig 7.
Spatial distribution of individual urban land cover disturbance indicators.
All thematic maps were produced by the authors.
Table 3.
Expert-based AHP judgment matrix for urban land-cover disturbance indicators.
Fig 8.
Spatial distribution of comprehensive urban land cover change intensity.
All visual elements were generated by the authors.
Table 4.
Land Cover Change Intensity at the Pixel and Regional Levels within the Study Area.
Table 5.
Model efficiency and performance comparative analysis.
Fig 9.
Large range of recognition effect comparison diagram in Langfang.
Background imagery was obtained from the Copernicus Data Space Ecosystem(Copernicus Sentinel data) and processed by the authors for visualization and comparison. The imagery corresponds to original satellite data and is not derived from any third-party online basemap or proprietary map service. All classification results, overlays, and map elements were generated by the authors.