Zhang et al. (2025) An Integrated Feature Framework for Wetland Mapping Using Multi-Source Imagery

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2025
• Date: 2025-11-17
• Authors: Liansong Zhang, Z.P. Wang, Jifei Wang, Qiang Hu, Yonglei Chang, Zhong Lu, Jinqi Zhao
• DOI: 10.3390/rs17223737

Research Groups

Not specified in the provided text.

Short Summary

This paper proposes an integrated framework combining knowledge-driven and data-driven features from multi-source imagery into a Random Forest classifier for wetland mapping, achieving superior classification performance, enhanced robustness, and improved interpretability across different study areas.

Objective

• To develop an integrated framework that combines knowledge-driven and data-driven features from multi-source imagery to form a complementary feature set for robust, accurate, and interpretable wetland mapping.

Study Configuration

• Spatial Scale: Regional (Yellow River Delta) and local (Qilihai Wetland) wetland ecosystems.
• Temporal Scale: Not specified in the provided text.

Methodology and Data

• Models used: Convolutional Neural Networks (CNNs) for data-driven feature extraction; knowledge-driven features derived from physical models and expert-defined indices; Random Forest (RF) classifier for final classification.
• Data sources: Multi-source imagery (specific types not detailed).

Main Results

• The proposed integrated approach achieved the best classification performance among all comparative experiments.
• For the Yellow River Delta: Overall Accuracy (OA) of 90.91%, Kappa coefficient of 0.8898, and F1-score of 0.9136.
• For the Qilihai Wetland: Overall Accuracy (OA) of 91.31%, Kappa coefficient of 0.8893, and F1-score of 0.9308.
• The integration of knowledge-driven and data-driven features effectively enhanced classification robustness and improved the interpretability of feature representations.

Contributions

• Proposes a novel integrated framework that synergistically combines knowledge-driven and data-driven features for wetland mapping, addressing the limitations of purely data-driven (overfitting, interpretability) and knowledge-driven (flexibility, adaptability) approaches.
• Demonstrates superior classification accuracy and robustness across diverse wetland environments (Yellow River Delta and Qilihai Wetland).
• Enhances model interpretability by visualizing and analyzing the importance of both knowledge-driven and data-driven features.

Funding

Not specified in the provided text.

Citation

@article{Zhang2025Integrated,
  author = {Zhang, Liansong and Wang, Z.P. and Wang, Jifei and Hu, Qiang and Chang, Yonglei and Lu, Zhong and Zhao, Jinqi},
  title = {An Integrated Feature Framework for Wetland Mapping Using Multi-Source Imagery},
  journal = {Remote Sensing},
  year = {2025},
  doi = {10.3390/rs17223737},
  url = {https://doi.org/10.3390/rs17223737}
}