Generative adversarial network with double discrimination for heterogeneous hyperspectral reconstruction

文献类型： 外文期刊

第一作者： Zhou, Haozhe

作者： Zhou, Haozhe;Shen, Yiyang;Yan, Ziyi;Zhang, Yanchao;Yang, Yongjie;Yu, Xiaoyue;Lu, Yongliang

作者机构：

关键词： Images registration; Hyperspectral reconstruction; Generation adversarial network (GAN); Spatial-spectral attention; UAV

期刊名称：COMPUTERS AND ELECTRONICS IN AGRICULTURE （ 影响因子：8.9； 五年影响因子：9.3 ）

ISSN： 0168-1699

年卷期： 2025 年 237 卷

页码：

收录情况： SCI

摘要： Both spatial and spectral resolution play important roles in the identification of objects in hyperspectral remote sensing. However, the imaging characteristics of hyperspectral imagery lead to a mutually constraining relationship between spatial and spectral resolution. With the development of DL, the inverse transformation based on HS reconstruction of multispectral image (MSI) images has made great progress. Previous work mainly utilized HSI downsampling to obtain a few bands as inputs to the model for reconstruction, which is merely image processing rather than practical application. In order to improve the accuracy and application of HS reconstruction in field production activities, we propose a generative adversarial network (GAN) based on double discrimination. Specifically, the dataset is first formed through the preprocessing process of image alignment; then, based on the proposed network with double discrimination, adversarial learning is performed from the spatial dimension and the spectral dimension to ensure the spectral similarity of the HSI reconstruction; and finally, we propose a VI-based discrimination method to ensure the practical application capability. The results show that the proposed method not only achieves good results in all commonly used evaluation metrics, but also has good capabilities in VI-based production activities and generalization capabilities. In conclusion, the proposed method can help researchers in the field of remote sensing to better guide their production activities with less input.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序