Shrivastava et al. (2026) Enhancing Land Classification Accuracy: A Comprehensive Study of Sentinel-1 and Sentinel-2 Image Fusion Techniques

Identification

• Journal: Lecture notes in networks and systems
• Year: 2026
• Date: 2026-01-01
• Authors: Priyanka Shrivastava, Mani Roja Edinburgh, Varsha Turkar
• DOI: 10.1007/978-3-032-10667-4_38

Research Groups

• Thadomal Shahani Engineering College, Mumbai, India
• Vidyalankar Institute of Technology, Mumbai, India

Short Summary

This study comprehensively evaluates various image fusion techniques combining Sentinel-1 SAR and Sentinel-2 optical data to enhance land classification accuracy. It demonstrates that integrating these complementary datasets significantly improves classification performance across agricultural, forestry, and ecological applications by leveraging the strengths of each sensor.

Objective

• To evaluate how image fusion of Sentinel-1 and Sentinel-2 can improve accuracy in land classification relevant to agricultural, forestry, and ecological management practices.
• To assess different data fusion methods (pixel, feature, and decision levels), study their application, and examine challenges and possibilities for advancing land classification techniques using machine learning algorithms.

Study Configuration

• Spatial Scale: Not explicitly defined for a specific study area, but the applications (water body mapping, forest burn detection, winter crop classification) imply regional to potentially national scale analyses.
• Temporal Scale: Not explicitly defined, but "winter crop classification" suggests seasonal relevance.

Methodology and Data

• Models used: Attention-based U-Nets, Feature-fusion CNNs, Long Short-Term Memory (LSTMs). Data fusion was applied at pixel, feature, and decision levels.
• Data sources: Synthetic Aperture Radar (SAR) data from Sentinel-1 and optical imagery from Sentinel-2.

Main Results

• Combining Sentinel-1 and Sentinel-2 data consistently led to more accurate land classification by balancing the strengths and weaknesses of each sensor.
• Specific application accuracies achieved:
  • Water body mapping: up to 99.38% accuracy.
  • Forest burn detection: a mean Intersection over Union (mIoU) of 88.3%.
  • Winter crop classification: exceeded 99% overall accuracy.
• The study found that advanced machine learning models, coupled with multi-level data fusion, provide clearer and more detailed insights for environmental monitoring.

Contributions

• Provides a comprehensive evaluation of Sentinel-1 and Sentinel-2 image fusion techniques for land classification, addressing the complementary nature of SAR and optical data.
• Demonstrates significant improvements in land classification accuracy across diverse applications (agriculture, forestry, ecology) using advanced machine learning algorithms and multi-level data fusion.
• Highlights the potential of integrated remote sensing data for enhanced environmental monitoring and management practices.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Shrivastava2026Enhancing,
  author = {Shrivastava, Priyanka and Edinburgh, Mani Roja and Turkar, Varsha},
  title = {Enhancing Land Classification Accuracy: A Comprehensive Study of Sentinel-1 and Sentinel-2 Image Fusion Techniques},
  journal = {Lecture notes in networks and systems},
  year = {2026},
  doi = {10.1007/978-3-032-10667-4_38},
  url = {https://doi.org/10.1007/978-3-032-10667-4_38}
}