Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FUSU: A Multi-temporal-source Land Use Change Segmentation Dataset for Fine-grained Urban Semantic Understanding

Shuai Yuan, Guancong Lin, Lixian Zhang, Runmin Dong, Jinxiao Zhang, Shuang Chen, Juepeng Zheng, Jie Wang, Haohuan Fu

TL;DR

This work tackles the lack of fine-grained urban land-use change data by introducing FUSU, a multi-temporal, multi-source dataset with 17 land-use classes and over $3\times10^{10}$ annotation pixels across $847\ \mathrm{km^2}$ in five Chinese cities, using bi-temporal high-resolution imagery ($0.2$–$0.5$ m) and monthly Sentinel-1/2 time series. To leverage this rich data, the authors propose FUSU-Net, a unified time-series architecture that jointly performs segmentation and change detection through a dual-branch design that fuses time-series Sentinel features with bi-temporal high-resolution features. Comprehensive experiments show that FUSU-Net achieves superior segmentation and change-detection performance, highlighting the benefits of multi-source, multi-temporal context for fine-grained urban understanding. The dataset and baseline model aim to advance multi-source, multi-temporal urban monitoring with practical implications for planning and environmental assessment, while also outlining limitations and directions for future multi-modal fusion and domain adaptation.

Abstract

Fine urban change segmentation using multi-temporal remote sensing images is essential for understanding human-environment interactions in urban areas. Although there have been advances in high-quality land cover datasets that reveal the physical features of urban landscapes, the lack of fine-grained land use datasets hinders a deeper understanding of how human activities are distributed across the landscape and the impact of these activities on the environment, thus constraining proper technique development. To address this, we introduce FUSU, the first fine-grained land use change segmentation dataset for Fine-grained Urban Semantic Understanding. FUSU features the most detailed land use classification system to date, with 17 classes and 30 billion pixels of annotations. It includes bi-temporal high-resolution satellite images with 0.2-0.5 m ground sample distance and monthly optical and radar satellite time series, covering 847 km^2 across five urban areas in the southern and northern of China with different geographical features. The fine-grained land use pixel-wise annotations and high spatial-temporal resolution data provide a robust foundation for developing proper deep learning models to provide contextual insights on human activities and urbanization. To fully leverage FUSU, we propose a unified time-series architecture for both change detection and segmentation. We benchmark FUSU on various methods for several tasks. Dataset and code are available at: https://github.com/yuanshuai0914/FUSU.

FUSU: A Multi-temporal-source Land Use Change Segmentation Dataset for Fine-grained Urban Semantic Understanding

TL;DR

This work tackles the lack of fine-grained urban land-use change data by introducing FUSU, a multi-temporal, multi-source dataset with 17 land-use classes and over annotation pixels across in five Chinese cities, using bi-temporal high-resolution imagery (– m) and monthly Sentinel-1/2 time series. To leverage this rich data, the authors propose FUSU-Net, a unified time-series architecture that jointly performs segmentation and change detection through a dual-branch design that fuses time-series Sentinel features with bi-temporal high-resolution features. Comprehensive experiments show that FUSU-Net achieves superior segmentation and change-detection performance, highlighting the benefits of multi-source, multi-temporal context for fine-grained urban understanding. The dataset and baseline model aim to advance multi-source, multi-temporal urban monitoring with practical implications for planning and environmental assessment, while also outlining limitations and directions for future multi-modal fusion and domain adaptation.

Abstract

Fine urban change segmentation using multi-temporal remote sensing images is essential for understanding human-environment interactions in urban areas. Although there have been advances in high-quality land cover datasets that reveal the physical features of urban landscapes, the lack of fine-grained land use datasets hinders a deeper understanding of how human activities are distributed across the landscape and the impact of these activities on the environment, thus constraining proper technique development. To address this, we introduce FUSU, the first fine-grained land use change segmentation dataset for Fine-grained Urban Semantic Understanding. FUSU features the most detailed land use classification system to date, with 17 classes and 30 billion pixels of annotations. It includes bi-temporal high-resolution satellite images with 0.2-0.5 m ground sample distance and monthly optical and radar satellite time series, covering 847 km^2 across five urban areas in the southern and northern of China with different geographical features. The fine-grained land use pixel-wise annotations and high spatial-temporal resolution data provide a robust foundation for developing proper deep learning models to provide contextual insights on human activities and urbanization. To fully leverage FUSU, we propose a unified time-series architecture for both change detection and segmentation. We benchmark FUSU on various methods for several tasks. Dataset and code are available at: https://github.com/yuanshuai0914/FUSU.
Paper Structure (26 sections, 4 equations, 8 figures, 13 tables)

This paper contains 26 sections, 4 equations, 8 figures, 13 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Urban Change Segmentation Data
  4. Remote sensing tasks
  5. FUSU Dataset
  6. Construction of FUSU
  7. Statistic
  8. FUSU-Net
  9. Preliminary and Overview
  10. Overall architecture
  11. Loss Functions
  12. Experiments
  13. Semantic Segmentation
  14. Change Detection
  15. Discussion
  16. ...and 11 more sections

Figures (8)

  • Figure 1: The unique features of urban areas. Compared with other geographic regions, urban areas have dense semantics, fast temporal changes, and involve a large amount of human activities.
  • Figure 2: The visualization of the FUSU dataset construction. Each patch has 27 images (25 Sentinel images and 2 high-resolution images), and 2 labels. The content of the high-resolution image is center-surrounded by the Sentinel image as the red rectangle shows.
  • Figure 3: The distribution of the FUSU dataset. (a) Xi'an and Jiaxing are located in different climate zones. (b) The 5 urban districts of Xi'an and Jiaxing in FUSU dataset. (c) The visualization of image samples.
  • Figure 4: The statistic of the FUSU dataset. (a) Pixels distribution. (b) Polygon distribution. (c) Class distribution of T1 Xi'an. (d) Class distribution of T1 Jiaxing. (e) Class distribution of T2 Xi'an. (f) Class distribution of T2 Jiaxing.
  • Figure 5: The architecture of FUSU-Net. (a) U-TAE branch for time-series images. (b) Bi-temporal branch for segmentation and change detection. (c) Feature fusion.
  • ...and 3 more figures