Hyperspectral and multispectral image fusion method and system and medium

Abstract

The invention discloses a hyperspectral and multispectral image fusion method and system and a medium. The hyperspectral and multispectral image fusion method comprises the following steps: inputtinga hyperspectral image Y and a multispectral image Z; extracting a spectral subspace S from the hyperspectral image Y by adopting singular value decomposition; estimating a subspace coefficient A according to the multispectral image Z and the obtained spectral subspace S; and multiplying the spectral subspace S and the subspace coefficient A to obtain a fused hyperspectral image X. According to themethod, the low-resolution hyperspectral image and the high-resolution multispectral image can be effectively fused to obtain the high-resolution hyperspectral image, and the method has the advantages of being high in reconstruction precision, high in calculation efficiency, high in universality and robustness and resistant to noise interference.

Description

technical field [0001] The invention relates to hyperspectral and multispectral image fusion technology, in particular to a hyperspectral and multispectral image fusion method, system and medium based on subspace representation and deep convolutional neural network. Background technique [0002] Hyperspectral imaging technology can obtain image information of dozens or hundreds of spectral bands at the same time, and rich spectral information is helpful for accurate identification of substances in the scene. Therefore, hyperspectral imaging technology is widely used in earth observation, military monitoring, Environmental monitoring, geological exploration, medical detection and face recognition and other fields. However, due to hardware limitations, it is difficult for existing imaging systems to directly obtain high-resolution hyperspectral images. Therefore, the spatial resolution of hyperspectral images is often very low, which greatly limits the application of hyperspe...

AI Technical Summary

Problems solved by technology

This type of method mainly has the following problems: firstly, the training data of high-resolution hyperspectral images does not exist in practice, and it is necessary to simulate the construction of the data set by downsampling the hyperspectral images, and training on the data of the simulation construction will be difficult. Reduce the accuracy of fusion; the second is that the number of spectral bands of hyperspectral images of different sensors is different, resulting in poor scalability of the training network, that is, a network trained in a class of hyperspectral images cannot be directly applied to high spectral image

Images