Scholarlink

In recent Hyperspectral image (HSI) classification research, Vision Transformer (ViT) has gained significant attention. However, ViT often struggles with the computational complexity of handling HSI patches and may not effectively emphasize crucial spectral bands. To tackle these challenges, this paper developes a novel HSI classification network called Spectral Channel-weighting Cross Attention in Vision Transformer (SCCAT). First, we integrate Cross Attention in Vision Transformer (CAT) into HSI classification, reducing computational requirements and achieving efficient classification. Second, we construct a Spectral Channel Weighting (SCW) module to address the network’s limited attention to spectral information. Finally, through extensive experiments on Indian Pines and Salinas datasets, we validate the superior classification performance of SCCAT in terms of accuracy.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Spectral Channel-Weighting CAT for Hyperspectral Image Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yujuan Qi,
  • Yuxuan Guo,
  • Baodi Liu,
  • Yanjiang Wang

摘要

In recent Hyperspectral image (HSI) classification research, Vision Transformer (ViT) has gained significant attention. However, ViT often struggles with the computational complexity of handling HSI patches and may not effectively emphasize crucial spectral bands. To tackle these challenges, this paper developes a novel HSI classification network called Spectral Channel-weighting Cross Attention in Vision Transformer (SCCAT). First, we integrate Cross Attention in Vision Transformer (CAT) into HSI classification, reducing computational requirements and achieving efficient classification. Second, we construct a Spectral Channel Weighting (SCW) module to address the network’s limited attention to spectral information. Finally, through extensive experiments on Indian Pines and Salinas datasets, we validate the superior classification performance of SCCAT in terms of accuracy.