Mapping Large-Scale Local Climate Zones From High-Resolution Multiview Satellite Imagery
Large-scale and reliable local climate zone (LCZ) information is crucial for the study of urban climate, such as urban heat island. Existing research for LCZ generation predominantly relies on planar information derived from single-view medium-resolution imagery, neglecting the 3-D attributes of sur...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10975142/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Large-scale and reliable local climate zone (LCZ) information is crucial for the study of urban climate, such as urban heat island. Existing research for LCZ generation predominantly relies on planar information derived from single-view medium-resolution imagery, neglecting the 3-D attributes of surface structures. In complex urban settings, there still remain big challenges for the accurate LCZ identification, especially for the LCZ types categorized according to height and density. High-resolution multiview (HRMV) satellite data hold great values for elaborate urban information extraction, but its potential to acquire LCZs has not been investigated yet. In this context, a novel method was proposed for LCZ mapping by taking full advantage of HRMV imagery. Our contributions are twofold: First, a set of multiangular indices (MAIs) was constructed to adequately depict the variation of angular disparities between HRMV images. Second, a planar–stereo feature fusion method was proposed for LCZ classification by jointly considering the spectral–spatial–angular information from HRMV data. To verify the effectiveness of the proposed method, Ziyuan-3 HRMV images over five large cities in China, namely Beijing, Chongqing, Guangzhou, Kunming, and Wuhan, were employed as the experimental data. Our results achieved an overall accuracy of 82.7% and <italic>F</italic>1-score of 0.81. The incorporation of the MAIs contributes significantly to accuracy improvement, especially to the building types (LCZs 1–10). Our findings demonstrate the advantages of HRMV satellite data for large-scale and fine-grained LCZ mapping, and offer new perspectives for future LCZ mapping tasks. |
|---|---|
| ISSN: | 1939-1404 2151-1535 |