资源一号02D卫星高光谱与多光谱遥感影像融合方法

Hyperspectral and Multispectral Remote Sensing Images Fusion Method of ZY-1-02D Satellite

资源一号02D卫星可以在同一时间同步获取30m空间分辨率的高光谱遥感影像和10m空间分辨率的多光谱遥感影像,对二者进行影像融合可以在保持高光谱影像光谱准确性的同时提升数据的空间细节特征,从而进一步提高卫星的应用价值。文章通过研究深度学习方法,构建了一个空谱特征分离式网络(Spatial-Spectral Features Separated Network,SSFSN),最终实现高光谱影像与多光谱影像的融合。为验证方法有效性,对资源一号02D卫星的高光谱影像和多光谱影像进行融合试验。实验结果表明:提出的方法在目视效果和质量评价指标上均优于对比方法。

ZY-1-02D satellite can simultaneously acquire hyperspectral and multispectral remote sensing images at 30-meter and 10-meter spatial resolution,respectively.Image fusion can improve the spatial detail characteristics of a hyperspectral image while maintaining its spectral accuracy,thus improving the application value of ZY-1-02D satellite.The paper presents a spatial-spectral features separated network(SSFSN)method based on deep learning to achieve hyperspectral and multispectral image fusion.In order to verify the effectiveness of the method,hyperspectral and multispectral images of ZY-1-02D satellite are utilized for the experiments.The experiments demonstrate that our method outperforms the methods compared,both in visual quality and evaluation indexes.