Related Experiment Videos
A Multi-Modal Remote Sensing Image Collaborative Fusion Network for Construction and Demolition Waste Extraction in
Ya-Zhe Xie1,2,3, Zhong-Qi Shi4,5, Lan-Qing Zhang6
1Key Laboratory of Digital Earth Science, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China.
Sensors (Basel, Switzerland)
|July 28, 2026
Summary
This study introduces a new dataset and AI model for mapping construction and demolition waste (C&DW) using satellite imagery. The approach improves accuracy in complex urban areas, aiding environmental management.
Area of Science:
- Remote Sensing
- Environmental Science
- Artificial Intelligence
Background:
- Accurate mapping of construction and demolition waste (C&DW) is vital for urban environmental management.
- Challenges in C&DW extraction include complex backgrounds, spectral confusion, and irregular boundaries in satellite imagery.
Purpose of the Study:
- To develop a novel optical-SAR multimodal semantic segmentation dataset for fine-scale C&DW extraction.
- To propose an effective deep learning framework (CDW-Net) for multimodal fusion in C&DW mapping.
Main Methods:
- Creation of a multimodal dataset using GF-3 SAR, Sentinel-1 SAR, and Sentinel-2 optical imagery at a 3m resolution.
- Development of CDW-Net, a dual-branch framework featuring a Cross-Frequency Interaction (CFI) module and a Multi-Source Boundary Attention Module (MS-BAM).
Main Results:
- CDW-Net achieved high performance in Beijing with an IoU of 0.8470 and F1-score of 0.9171, surpassing existing methods.
- Demonstrated strong cross-regional transferability with an IoU of 0.8247 and F1-score of 0.9040 in Guangzhou.
Conclusions:
- The proposed dataset and CDW-Net framework significantly advance fine-scale C&DW monitoring capabilities in complex urban settings.
- This work offers valuable tools for urban environmental management and sustainable development through improved waste mapping.