DCN-T: Dual Context Network With Transformer for Hyperspectral Image Classification.

Area of Science:

• Computer Vision
• Remote Sensing
• Machine Learning

Background:

• Hyperspectral image (HSI) classification faces challenges due to spatial variability and limited annotated data.
• Existing methods often have limited representation ability due to training networks from scratch.

Purpose of the Study:

• To develop an effective HSI classification method overcoming limitations of prior approaches.
• To leverage pre-trained networks for improved feature extraction from HSI data.

Main Methods:

• A tri-spectral image generation pipeline transforms HSI into high-quality tri-spectral images.
• An end-to-end segmentation network (DCN-T) using transformers encodes spatial contexts within homogeneous areas.
• An ensemble approach integrates segmentation results from tri-spectral images via a voting scheme.

Main Results:

• The proposed method achieves superior performance on three public benchmarks for HSI classification.
• Outperforms state-of-the-art methods in hyperspectral image classification tasks.

Conclusions:

• The novel pipeline and DCN-T network effectively address HSI classification challenges.
• The method demonstrates significant improvements in accuracy and representation ability.