A new band selection framework for hyperspectral remote sensing image classification
2024

New Framework for Hyperspectral Image Classification

publication Evidence: high

Author Information

Author(s): Phaneendra Kumar B. L. N., Vaddi Radhesyam, Manoharan Prabukumar, Agilandeeswari L., Sangeetha V.

Primary Institution: Vellore Institute of Technology, Vellore

Hypothesis

Can a new band selection framework improve classification accuracy in hyperspectral remote sensing images?

Conclusion

The proposed method achieved high classification accuracy of 99.92% on Indian Pines, 99.94% on Salinas, and 97.23% on KSC datasets.

Supporting Evidence

  • The proposed method reported overall accuracy of 99.92% on Indian Pines dataset.
  • The method achieved 99.94% accuracy on Salinas dataset.
  • 97.23% accuracy was reported on the KSC dataset.
  • Mean Spectral Divergence values were 42.4, 63.75, and 41.2 for the three datasets respectively.

Takeaway

This study created a new way to pick the best parts of images taken from space, helping to identify different crops very accurately.

Methodology

The study used a dual partitioning strategy for band selection and a Convolutional Neural Network for classification.

Limitations

The method requires manual tuning of parameters, which may affect its adaptability.

Digital Object Identifier (DOI)

10.1038/s41598-024-83118-8

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication